Beyond Kolmogorov Cascades

Downloaded from https://www.cambridge.org/core.
University of Strathclyde, on 19 Jan 2022 at 08:33:08, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2019.98
J. Fluid Mech. (2019), vol. 867, P1, doi:10.1017/jfm.2019.98
cambridge.org/jfmperspectives
Beyond Kolmogorov cascades
Bérengère Dubrulle†
SPEC, CEA, CNRS, Université Paris-Saclay, F-91191 CEA Saclay, Gif-sur-Yvette, France
The large-scale structure of many turbulent flows encountered in practical situations

such as aeronautics, industry, meteorology is nowadays successfully computed using
the Kolmogorov–Kármán–Howarth energy cascade picture. This theory appears
increasingly inaccurate when going down the energy cascade that terminates
through intermittent spots of energy dissipation, at variance with the assumed
homogeneity. This is problematic for the modelling of all processes that depend on
small scales of turbulence, such as combustion instabilities or droplet atomization in
industrial burners or cloud formation. This paper explores a paradigm shift where
the homogeneity hypothesis is replaced by the assumption that turbulence contains
singularities, as suggested by Onsager. This paradigm leads to a weak formulation
of the Kolmogorov–Kármán–Howarth–Monin equation (WKHE) that allows taking
into account explicitly the presence of singularities and their impact on the energy
transfer and dissipation. It provides a local in scale, space and time description
of energy transfers and dissipation, valid for any inhomogeneous, anisotropic flow,
under any type of boundary conditions. The goal of this article is to discuss WKHE
as a tool to get a new description of energy cascades and dissipation that goes
beyond Kolmogorov and allows the description of small-scale intermittency. It puts
the problem of intermittency and dissipation in turbulence into a modern framework,
compatible with recent mathematical advances on the proof of Onsager’s conjecture.
Key words: intermittency, turbulence theory
‘I soon understood that there was little hope of developing a pure,

closed theory, and, because of absence of such a theory, the investigation
must be based on hypotheses obtained on processing experimental data’.
[Kolmogorov, quoted by A. Tsinober]
† Email address for correspondence: berengere.dubrulle@cea.fr
c Cambridge University Press 2019
867 P1-1
Downloaded from https://www.cambridge.org/core. University of Strathclyde, on 19 Jan 2022 at 08:33:08, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2019.98
B. Dubrulle
1. Introduction
The first lesson I learned as a graduate student is that there is no official
definition of turbulence. According to consensual observations, turbulence is a
state of fluids characterized by swirling motions. Such motions are observed in a
variety of phenomena, covering a wide range of scales, from the centimetre tap
water vortex to 102 km size hurricanes or 104 light year galaxies. The building
of a comprehensive theory of turbulence, allowing understanding and prediction of
these phenomena, is therefore a long-standing issue. The situation of turbulence
with respect to other outstanding problems in physics, such as the unification of
gravity and quantum mechanics, is paradoxical. Contrary to the other problems, the
constitutive equations governing the dynamics of turbulent flows are well-known.
They are the incompressible Navier–Stokes equations (INSE) for unstratified fluids:
1
∂t ui + uj ∂j ui = − ∂i p + ν∂j ∂j ui + fi , (1.1)
ρ
∂j uj = 0, (1.2)
where ui is the d-dimensional velocity field, p the kinematic pressure, ρ the
(constant) density, fi is the forcing and ν the molecular viscosity. Since Reynolds’
pioneering work, we know that after a proper rescaling of the equations by L and
U, some characteristic length and velocity, the INSE only depend on one parameter,
the Reynolds number Re = LU/ν.
Equipped with a set of equations depending only of one parameter, we should
be in the most favourable situation to build a theory of turbulence – at least,
that is often what my colleagues in quantum mechanics seem to think. When
confronted with this question, I always give the same answer, which is probably the
second lesson I learned from my turbulence classes: the INSE are nonlinear partial
differential equations, and they have been resisting mathematical and analytical
treatments for almost two centuries: (i) it is still not known whether their solutions
with finite energy remain regular for all time – this is the subject of the Clay
prize – and (ii) very few analytical solutions exists. Until the first half of the 20th
century, before the advent of computers, any theory of turbulence could therefore
only be based on hypotheses derived from observations or experimental data, as
noted by Kolmogorov. Using a minimal set of such hypotheses, he was able in
1941 to construct a simple theory of turbulence based on the concept of energy
cascade. Most of the present models or theories of turbulence are based on the
Kolmogorov picture. It appears to describe successfully the large-scale structure
of many turbulent flows encountered in practical situations such as aeronautics,
industry, meteorology or astrophysics, but becomes increasingly inaccurate when
going down the scales. This is problematic for the modelling of all processes that
depend on small scales of turbulence, such as combustion instabilities, droplet
atomization in industrial burners or cloud formation.
867 P1-2
Beyond K41
We now have powerful computers, insist my quantum physicist colleagues. Why
not abandon any modelling and turn to direct numerical simulation (DNS) of
turbulent flows using the INSE? My reply to this is of course that DNS implies
discretizing the equations in space and time, i.e. choosing an appropriate time
and space resolution. For this, we still need a good model of turbulence. If we
use Kolmogorov theory, we find that to simulate a flow of scale L and velocity
U at a given Reynolds number Re, the resolutions in space and time should
be 1x ∼ LRe−3/4 and 1t ∼ (L/U)Re−1/2 . The number of nodes N to simulate then
scales like Re9/4 and the number of time steps to reach a time T = L/U is Re1/2 ,
making the computational burden O(Re11/4 ) to simulate the flow over such a period.
For any aeronautical, geophysical or astrophysical application, this exceeds by far
the capabilities of the most powerful computers. It seems that, for the time being,
we still are in need of a theory of turbulence that goes beyond the Kolmogorov
picture so as to be able to handle better the small scales of turbulence and provide
us with as sharp as possible bounds on the time and space discretization of DNS.
These issues have been at the heart of my research for over thirty years. Thanks
to general scientific progress, we now have better clues regarding mathematical
or analytical properties of INSE with respect to the Kolmogorov era, and better
mathematical tools to handle them. Most of all, thanks to progress in imaging and
computers, we have access to a much more detailed set of observations and data.
In this essay, I therefore restart Kolmogorov’s approach to turbulence from scratch,
and investigate the physics of turbulence by processing experimental and numerical
data. Combining the results with recent advances in mathematics, I show that it is
possible to go beyond Kolmogorov’s simple cascade picture and provide a modern
picture of turbulence energetics and scaling. This new picture can be used to derive
new prospects for the theory of turbulence, to be hopefully solved by the new
generation of turbulence researchers.
The numerical data will be based on DNS of INSE at moderate space resolution,
so as to be able to accumulate time statistics. They are described in Debue
et al. (2018). The experimental data sets are based on torques and velocity fields
measured in the von Kármán flow, using torquemeters and stereoscopic particle
image velocimetry at different resolutions and Reynolds numbers. These data are
described in Ravelet, Chiffaudel & Daviaud (2008), Saint-Michel et al. (2014). The
main characteristics of these data are summarized in table 1. The geometry of the
von Kármán flow is depicted in figure 1. The numerical data are by construction
noise free and representative of homogeneous isotropic turbulence, but are limited in
Reynolds number, and space and time statistics. The experimental data are subject
to unavoidable noise and include anisotropy and inhomogeneities, but cover a wide
range of scales and Reynolds numbers, with good statistics. In this article, I will
mainly use the experimental data, but resort to numerical data to remove any doubt
regarding the universality or validity of a given result.
867 P1-3
B. Dubrulle
Case F Glycerol Re Rλ η 1x Symbol

(Hz) content (adim) (mm) (mm)
ANTIG 5 0% 3 × 105 1870 0.045 0.02 2.4 E
ANTIC-1 5 0% 3 × 105 2750 0.0450 0.02 0.48 @
ANTIC-2 5 0% 3 × 105 2510 0.045 0.02 0.24 6
ANTIC-3 1 0% 4 × 104 917 0.045 0.08 0.48 A
ANTIC-4 1.2 59 % 6 × 103 214 0.045 0.37 0.24 ?
CONC-3D 0.1 0% 6 × 103 — 0.019 0.37 0.96 —
DNS A — 5123 — 138 0.02 0.009 kmax η = 1.6 ——
DNS B — 7283 — 54 0.11 0.032 kmax η = 4.2 ——
DNS C — 7283 — 146 0.08 0.009 kmax η = 1.1 ——
TABLE 1. Parameters describing the main datasets used in this paper. ANTI (respectively
CON) means impeller rotation in the scooping (respectively unscooping) direction, while
G or C refers to the location of the measurement (G for global, C for centre, see figure 1;
F is the rotation frequency of the impellers in Hz; Re is the Reynolds number based on
the radius of the tank; Reλ is the Taylor-microscale Reynolds number; is the global
dimensionless energy dissipation; η is the Kolmogorov dissipation length scale; and 1x
represents the spatial resolution in the measurements and the DNS. The last column
shows the symbols used to represent the experimental datasets. The experiment CONC-3D
corresponds to a case where the velocity is measured by tomographic particle image
velocimetry, providing access to the 3 components of velocity.
For the sake of clarity, I kept the calculations in the present paper to a minimal
complexity level. I refer the reader to the lecture notes of Eyink (2007–2008), where
many details about Onsager’s conjecture, multi-fractal theory and weak solutions can
be found.
2. Energetics of turbulence and the cascade picture
2.1. Global energy budget
In 1845, James Prescott Joule published an account of his experiment showing that
the application of a 772.24 foot pound force (≈1050 J) on a paddle stirring a pound
of water raises the temperature of the fluid by one degree Fahrenheit. This is an
illustration of a global process occurring in most fluids, by which mechanical energy
(work) is converted into thermal energy (heat), in agreement with the first law of
thermodynamics governing the variation of the fluid total energy in a closed system
1Etot = Q + W, as a function of the heat Q and work W added. The actual fluid
energy balance of Joule’s experiment is easily derived from the constitutive equation
of the fluid. Taking the scalar product of (1.1) with u and integrating over the whole
volume, we then get R the equation governing the time variation of the total kinetic
energy E = (1/2)∂t ρu2 d x
Z Z
∂t E = ρu · f d x − ν ρ (∇u)2 d x (2.1)
≡ Pinj − Pdiss , (2.2)
867 P1-4
Beyond K41
(a) (b) f2, C2
1.4 1.8
rπ
f1, C1
0.925 1=R
(c)
F IGURE 1. The von Kármán flow geometry. (a) General view of the experiment in
Saclay. The von Kármán flow is generated in a Plexiglas cylinder of radius R, filled with
fluid at a controlled temperature. The flow is driven by two counter-rotating curve-bladed
impellers at frequencies f1 = f2 , rotating with their blades concave face pushing forward,
corresponding to the ‘–’ sign in (c) (hereafter referred to as ANTI). For a given F =
( f1 + f2 )/2, this forcing produces the most intense turbulence, with strongest fluctuations.
(b) Schematic design of the experiment. The torque and frequency of each impeller
is measured using a torquemeter. Unless explicitly specified, all experimental quantities
presented in the present paper are made dimensionless using R as a unit of length, and
1/(2πF) as a unit of time. The mean velocity field measured through stereoscopic particle
image velocimetry (SPIV) is shown as an inset. The arrows trace the in-plane velocity.
The colour traces the out-of-plane velocity. Three types of measurement were done. Either
over the whole plane in between the impeller, corresponding to the area covered by the
inset (referred to as G for global); or inside the black square (referred as C for centre);
or inside the red square (referred to as W for wall). Most of the measurements used in
the present essay are done with impellers rotating in the − direction and performed in
C where the turbulence is the most isotropic and homogeneous. In some cases, we have
also added measurements with + direction (called CON), and in the W area, to explore
the effects of inhomogeneity, anisotropy or forcing. Adapted from Debue (2019).
where Pinj is the injected power through the mechanical work of the paddle, and
Pdiss is the dissipated energy through the action of the viscous forces. An example
of such balance computed in the statistically stationary DNS A of table 1 is shown
in figure 2. Besides checking the validity of (2.2), this figure illustrates a puzzling
feature: the injected power and the dissipated power display similar oscillations
around a common mean value Pinj = Pdiss , albeit shifted by a constant time lag τc .
867 P1-5
B. Dubrulle
0.05
dE/dt, Pinj, Pdiss, (Pinj - Pdiss)

0.04
0.03
0.02
0.01
-0.01
-0.02
0 20 40 60 80 100 120 140 160 180
t
F IGURE 2. Global energy budget for DNS A. The different terms of (2.2) are plotted as
a function of (non-dimensional) time: Pinj : blue continuous line; Pdiss : red dotted line; ∂t E:
black dotted line. The green dotted line is Pinj − Pdiss , which should coincide with ∂t E
in a stationary state. This is indeed the case for this DNS where both curves superpose
exactly.
This means that the conversion of work into heat is not instantaneous nor direct,
and takes place within the flow via a process that connects different parts of the
fluids. To unravel such a process, it is necessary to analyse velocity correlations
inside the fluids.
2.2. Velocity correlations and energy spectra

A classical statistical measure of spatial velocity correlations is the function
C(x, r, t) = hu(x, t)u(x + r, t)i, where h·i means statistical average. For any practical
application throughout this paper, we shall take advantage, whenever appropriate, of
statistical homogeneity, stationarity or isotropy to estimate it through spatial average
(hAix ≡ hAi), time average (hAit ≡ A) or an angle average over the scale direction
(hAiφ ).
If the turbulence is statistically homogeneous and stationary, C(x, r, t) depends
neither Ron space x nor on time t, and can be studied through its Fourier transform,
E (k) = eik·r dr. If, further, the turbulence is isotropic, E (k) only depends on k = kkk
R the energy spectrum E(k) = Ak E (k), where A is a normalization factor
d−1
and defines
so that E(k) dk = 0.5hu(x, t)u(x, t)i.
The energy spectrum corresponding to DNS A is shown in figure 3(a). It is
self-similar at small wavenumbers (large scales), with a power law approximately
equal to k−5/3 and exponentially decays at large wavenumbers (small scales).
The spectrum peaks at a wavenumber kf = 1/Lf corresponding to the forcing
scale Lf . The exponential decay starts around η = (ν 3 /)1/4 . This is actually the
only characteristic scale one can build using the fluid viscosity ν and the mean
867 P1-6
Beyond K41
(a) 105 (b)

E(k)/´2/3˙5/3 104
100
102
10-5 100
10-10 10-2
10-2 100 10-2 100
k˙ k˙
F IGURE 3. Spatial kinetic energy spectra at various Rλ , and for different forcing or
anisotropy conditions. The black dotted line is k−5/3 . (a) In numerical simulations: DNS B:
blue line; DNS C: red line; the black continuous line is from the John Hopkins University
Turbulent database (http://turbulence.pha.jhu.edu), at Rλ = 433 (Li et al. 2008). Picture
courtesy F. Nguyen and J-Ph. Laval. (b) In experiments: CON C: red circle, CON W:
yellow circle; ANTI C: blue square; ANTI W: green square. Adapted from Debue (2019).
injected power per unit mass = Pinj /ρ V where V is the volume enclosing the fluid.
If we assume that the shape of the spectrum only depends on two variables, and
ν, we can further build an energy by unit wavenumber 2/3 η5/3 , so that the energy
spectrum is characterized by the function E(k)/ 2/3 η5/3 as a function of kη, which
are the ‘universal coordinates’ considered in figure 3(a). Is this indeed a ‘universal
representation’, i.e. does it change when we vary the parameters of the flow?
2.3. Universality of the spectrum

The first thing we can change is the Reynolds number. In figure 3(a), we add
the energy spectrum of two additional DNSs, one at lower and one at higher
Reynolds number. We observe that the range over which the self-similar range
is observed is decreased when the Reynolds is decreased, but apart from that, no
noticeable change is observed, and the three spectra collapse in universal coordinates.
Increasing further the Reynolds number is difficult through numerical simulations.
We can turn to experiment to reach larger Reynolds numbers, but then the isotropy
and homogeneity conditions will not be met anymore, at least not as cleanly as in
numerical simulations. This is anyway a good test of robustness of the universality
spectrum.
To illustrate this, we plot in figure 3(b) the spatial energy spectrum, computed
for a series of experimental data fields measured by SPIV in a turbulent von
Kármán flow, at different spatial locations, using a variety of forcing and flow
parameters. The von Kármán flow being globally non-homogeneous, we cannot
expect the statistics of the turbulence to be identical everywhere in the tank (Debue
et al. 2018). We have selected two special locations in that respect (see figure 1):
867 P1-7
B. Dubrulle
(i) an area centred around the centre of the experimental set-up, where the mean
velocity is zero, and where large velocity fluctuations exist, due to the presence of
a strong mixing shear layer; (ii) an area centred around the mid-plane, but located
near one of the edges of the tank. At such location, turbulence is highly anisotropic,
and dominated by boundary layer effects. Measurements corresponding to these
different locations have been associated with different colours in figure 3(b). We
see no systematic differences with respect to the numerical energy spectrum.
The flow forcing conditions and the flows properties have also been changed
by changing the rotation direction of the impellers (labelled by different symbols
in figure 3b), as well as the flow viscosity. This results in different values of
(measured independently by torquemeters), Rλ (computed using the root-mean-square
of the velocity and the viscosity) and η (computed from the value of the viscosity
and ) for different data sets. Again, we observe that all spectra collapse on the
same universal curve as the DNS. This means that properties of the function
E(kη)/ 2/3 η5/3 are very robust, an indication that it is built in the properties of
the equations of motions. It was indeed Kolmogorov’s remarkable achievement to
derive a simple explanation for the shape of E(k) from simple manipulations of
the INSE, resulting in what is commonly called the ‘Kolmogorov 1941 theory of
turbulence’ (hereafter K41, (Kolmogorov 1941)).
3. The K41 theory and its consequences

3.1. Summary of K41 theory
The explanation of Kolmogorov is based on the Kármán–Howarth–Monin equation
(hereafter KHM), an equation that is derived from the Navier–Stokes equations by
taking their second moment and assuming local homogeneity (see e.g. Frisch (1996)
for an elementary derivation). The result is a dynamic equation for the correlation
function
1 1 ν
∂t C(r, t) − P` = ∇ r · hδu(δu)2 i − ∇r2 h(δu)2 i ≡ ∇ r J`
2 4 2
ν
≡ −Π` − Π` ,
I
(3.1)
where hi denotes the statistical average, ∇ r is the divergence with respect to r, ` =
krk denotes the scale, P` = hu(x) · (f (x + r) + f (x − r))i/2 is a measure of the mean
energy injection rate at scale `, δu = u(x + r) − u(x) is the velocity increments over
a displacement r.
For very large value of `, P` → 0 since the force and the velocity become
uncorrelated. For ` → 0, P` tends to , the mean energy injection rate (which is
also the mean energy dissipation rate). More precisely, for ` Lf , where Lf is
the forcing scale, one can Taylor expand f (x + r) and f (x − r) to second order in `
to get P` = + (O(`/Lf )2 ). Now, we note that the last term of the right-hand side
of (3.1) is proportional to the viscosity. Kolmogorov first assumes that there is a
867 P1-8
Beyond K41
range of scales Lmin ` Lf – the ‘inertial range’, to be characterized consistently
later by providing expression of Lmin – where this term is negligible with respect to
the first term of the right-hand side. Looking for a statistically stationary solution,
he then takes the time average to get ∇ r ·hδu(δu)2 i = −4, which can be readily
integrated into
h(δu)3 i = −4r/3. (3.2)
This is called the ‘Kolmogorov 4/3 law’ and is a basic law of turbulence.
Kolmogorov further assumes that, over the range of scales where this holds,
3/2
the function S = h(δu)3 i/h(δu)2 i is constant. We shall come back to this
assumption later. It means anyway that h(δu)2 i ∝ (`)2/3 , with ` = krk, or, after
twice differentiating ∇r2 C(r) ∝ 2/3 `−4/3 , since h(δu)2 i = 2(hu2 i − C(r)). By Fourier
transform, we then get k2 E(k) ∝ 2/3 k1/3 , i.e. E(k)/ 2/3 η5/3 ∝ (kη)−5/3 , which is the
behaviour we observe for small enough kη in figure 3. Moreover, the scaling of
C(r) provides us with an estimate of the range of the inertial range. It is indeed
characterized by scales ` ∼ 1/k such that:
∇ r · hδu(δu)2 i ∼ ν∇r2 h(δu)2 i ∼ ν 2/3 `−4/3 , (3.3)
which is valid as long as ` (ν 3 /)1/4 = η, i.e. kη 1. This is indeed what is
observed in figure 3.
3.2. K41 as a phenomenology of cascade and energy balance in scale space

The theory of Kolmogorov successfully explains, as we have seen, the shape of
the energy spectrum in universal coordinates in the inertial range. From a broader
point of view, it can in fact be seen as a theory of energy balance in the scale-time
domain, that helps to build a phenomenology of the energy cascade from the
injection scale to the dissipation scales. Indeed, comparing (3.1) with (2.2), we
see that KHM describes a global energy balance for C(r, t) that can be seen as
representative of the energy contained at scale ` = krk. Its time variation is governed
by two types of terms: the first one is the divergence, in the scale space, of a vector
J` = (1/4) · hδu(δu)2 i − ν∇ r h(δu)2 i/2. The second term is independent of scale and
describes the energy injection/dissipation . The vector J` can be split into two
components, one ‘inertial’ JI` = (1/4)hδu(δu)2 i, that only depends on velocity, and
a viscous component Jν` = −ν∇ r h(δu)2 i/2. The vector J` encodes an energy flux
through the scale space (rate at which energy is transferred from scales to scales).
The theory of Kolmogorov already provides us with the scaling of the divergence
of the time average of the two components in the inertial range

ΠÌ ≡ −∇ r · JI` ∼ , 

`
−4/3 (3.4)
Π`ν ≡ −∇ r · Jν` ∼ ν 2/3 `−4/3 ∼ ,
η

867 P1-9
B. Dubrulle
which is valid as long as ` > η. At ` ∼ η, the two terms coincide. The typical velocity
at this scales is uη ∼ (ηJI` )1/3 ∼ (η)1/3 , so that the Reynolds number at this scale is
Re(η) = (η4 )1/3 /ν = 1: viscosity becomes the preponderant process and the flow is
laminar below η. It is then meaningful to perform a Taylor expansion on the velocity,
√
to get the scaling of the velocity increment as δu ∼ r · ∇u = O(` /ν), where we
have used ∼ ν(∇u)2 . We thus infer the following scaling laws for scales below η
as
3/2 2 
`
Π` ∼ ` ∼
2
,
I

ν η

(3.5)
ν

Π` ∼ ν ∼ .


ν

The viscous component then carries most of the flux, and the inertial component
vanishes like `2 . The different scaling laws are summarized in figure 4(a). The
overall picture associated with these scalings is that of an ‘energy cascade’: the
energy is injected at large scale at a rate . It is transferred down the scales at a
constant rate by the inertial flux, until it reaches the scale η where the viscous
flux takes over, and carries this flux down to ` ∼ 0 (actually the hydrodynamic scale)
where it is transformed into heat. The time lag τc observed in figure 2 between
the time of energy injection and energy dissipation corresponds then to a ‘cascade
time’, the time it takes for a parcel of energy to be transferred from scale Lf to η.
The validity of this picture can be tested against DNS of homogeneous, isotropic
turbulence. This is done in figure 5(a). One sees that all the scalings laws of the K41
picture of turbulence are satisfied. Moreover, the ‘inertial’ range indeed expands as
Lf /η.
3.3. Universality of the energy budget in scale

The energy budget in the scale given by (3.1) and its associated picture of the energy
cascade (figure 4a,b), derived for ideal, homogeneous turbulence, is actually very
robust. Not surprisingly (remember the results for the spectra), the scaling properties
of the divergence of the time average of the two components of the energy flux
density J` , reported in figure 5(b), display a remarkable agreement with the expected
scaling laws, with no systematic effect due to different locations (coded by colour),
or forcing condition (coded by symbols). The main difference with the results of the
DNS is the value of the plateau for ΠÌ , which is around 0.3 rather than 1. This is
because, due to the planar measurement, we can only have access to the in-plane
component of this flux. If the flow were isotropic, we would thus get 2/3 of the
total value. The von Kármán flow is highly anisotropic at this location, so that the
out-of-plane component actually carries a large fraction of the flux, which explains
our result.
867 P1-10
Beyond K41
(a) (b)
100 ¶3h-1
100
Ô I¶/´, Ô ˜¶/´
DI¶/´, D˜¶/´
10-1
10-1
¶-4/3 ¶2h-2
¶2 ¶2
10-2 10-2
10-3 -2 10-3
10 100 102 10-2 100 102
¶/˙ ¶/˙
F IGURE 4. K41 versus multi-fractal picture of turbulence. (a) K41 picture: the energy
injected at scale Lf (vertical black dashed line) is transferred at a constant rate by
the energy flux ΠÌ (red line) down the scale where it becomes equal to the viscous
flux Π`ν (blue line). From downwards on, it is transported by the latter to the smallest
hydrodynamic scale, where it is dissipated into heat. In the inertial range, the energy
flux is constant, and the viscous flux increases like `−4/3 . In the dissipative range, the
viscous flux is constant, while the energy flux decreases like `2 in the dissipative range.
The energy flux and the viscous flux become equal at the Kolmogorov scale, indicated
by a black dotted line. (b) Multi-fractal picture: the energy is transferred at smaller scale
by the energy flux (red line) up to the scale ηh (vertical dashed dotted line) where the
energy flux and the viscous flux (blue line) balance. In the inertial range, the energy flux
and viscous flux scale respectively like `3h−1 and `2h−2 , while they obey the same scaling
than in the K41 picture in the dissipative range. When h < 1/3, the energy flux and the
viscous flux become equal at a scale smaller than the Kolmogorov scale, indicated by a
black dotted line.
3.4. Statistical properties of the energy cascade in scale space

Experiments provide large data sets of independent realizations of the cascade so
that we can use them to study the statistical properties of the energy cascade in scale
space. In figure 6(a,b), we plot the probability distribution functions (PDFs) of ΠÌ
and Π`ν , obtained through velocity measurements at uncorrelated times at the centre
of the experiment, for a flow forced in the scooping direction (minus sign direction
in figure 1(c)), for scales ranging in 1 6 `/η 6 103 . We see that the two PDFs are
very different and vary significantly with scale: the PDF of ΠÌ is almost symmetrical
around 0 with exponential tails widening as the scale increases; the PDFs of Π`ν are
lognormal (parabolic in log–log plot, not shown) with tails narrowing as the scale
increases.
Both PDFs however carry the same universal features: when plotted using centred
reduced variables ΠÌ∗ = (ΠÌ − ΠÌ )/σ I and Π`ν∗ = (Π`ν − Π`ν )/σ ν , the PDFs collapse
on two universal curves, see figure 6(a,b). The PDF of Π`ν∗ is similar in shape to the
PDF of energy dissipation (displayed in figure 6b1). Summarizing, we see that both
the local energy transfer ΠÌ and dissipation Π`ν have universal statistical properties,
867 P1-11
B. Dubrulle
(a) (b)
100 100
¶-4/3 ¶2
¶2
Ô I¶/´, Ô ˜¶/´
10-1
¶-4/3
10-2
10-2
10-3 10-4
100 102 100 102
¶/˙ ¶/˙
F IGURE 5. Test of K41 picture. (a) in numerical simulations: left-pointing triangles: DNS
B; right-pointing triangles: DNS A. Red symbols: mean energy flux ΠÌ /; blue symbol:
mean viscous flux Π`ν /. The red dotted line is 0.1(`/η)2 and the blue dashed line is
5(`/η)−4/3 . (b) In experiments using SPIV data: mean energy flux for CON C: red circle,
CON W: yellow circle; ANTI C: red square, ANTI C: yellow square; mean viscous flux
for CON C: blue circle, CON W: green circle, ANTI C: blue square, ANTI W: green
square. Adapted from Debue (2019).
following
!
ΠÌ − ΠÌ
P(ΠÌ ) = P I , (3.6)
σ I (`)
Π`ν − Π`ν

P(Π`ν ) = P ν , (3.7)
σ ν (`)
where all the scale dependences are encoded in the means ΠÌ , Π`ν and the standard
deviations (std) σ I (`), σ ν (`) shown in figure 7.
Not surprisingly, the means of Π`ν and ΠÌ obey the Kolmogorov scaling sketched
in figure 4(a). The standard deviations however, display new behaviours that are
not predicted by K41 theory: the std of Π`ν decreases with increasing scale with
an exponent slightly smaller than −4/3, while the std of ΠÌ increases slightly
with scales (like `0.25 ). To understand such behaviour, one needs to leave empirical
observation and dig deeper into theoretical consequences and meaning of K41.
4. K41 theory and symmetries: successes and limitations

4.1. Some basic symmetries of Navier–Stokes
As pointed out by Frisch (1996), K41 theory is deeply connected with basic
symmetry properties of the Navier–Stokes equations (1.1), namely:
(i) space translation (homogeneity): x → x + r, for arbitrary r;
(ii) time translation (stationarity): t → t + τ , for arbitrary τ ;
867 P1-12
Beyond K41
(a) log (¶/˙) (b) 0 (b1)

0
log10(P)
-2
3.0
0 -4
-6
2.5
-1 -8
-2 -2 0 2 4 6
2.0 ´*
log10(P)
-2 -4
1.5
1.0 -6
-3
0.5
-8
-4
-10 0 10 -2 0 2 4 6
Ô¶I* Ô¶˜*
F IGURE 6. PDFs of flux terms, centred and reduced as explained in the text, for
experiments ANTIG and ANTIC-1 to ANTIC-4 (see figure 7 for corresponding mean
and standard deviation and table 1 for symbols). (a) Centred and reduced PDF of ΠÌ at
different scale, coded by colour, following the colour bar. (b) Centred and reduced PDF of
Π`ν at different scale, coded by colour. Inset (b1): centred and reduced PDF of dissipation.
(a) (b)
100
¶1/4
100
Ô I¶/´, Ô ˜¶/´
ß I¶/´, ß˜¶/´
10-2
10-2 ¶-7/6
¶-4/3
100 101 102 103 100 102

¶/˙ ¶/˙
F IGURE 7. Scale variation of the mean and standard deviations of the flux terms for
experiments ANTIG and ANTIC-1 to ANTIC-4. (a) Time average: ΠÌ : red symbols; Π`ν :
blue symbols. (b) Standard deviation σÌ : red symbols; σ`ν : blue symbols. The black dashed
line is `1/4 . The dashed-dot line in (b) is `−4/3 `1/6 deduced from the `2 scaling of the std
ratio shown in figure 8(a) and the scaling of σÌ .
(iii) rescaling: (t, x, u) → (λ2 t, λx, λ−1 u), for arbitrary λ; and in the inviscid limit,
ν → 0:
(iv) time reversal: (t, u) → (−t, −u),
(v) h rescaling: (t, x, u) → (λ1−h t, λx, λh u), for arbitrary λ and h. Note that the
rescaling (iii) is equivalent to a −1 rescaling (property (v) with h = −1).
Let us explore in more details such connections and their consequences.
867 P1-13
B. Dubrulle
(a) (b) 0.6

Ô I/Ô˜3/2, ß I/(ß ˜)3/2 0.5
105 ¶2
0.4
Std(∂W)/¯∂W˘
0.3
100 ¶-0.095
0.2
100 102 100 102

¶/˙ ¶/˙
F IGURE 8. Test of rescaling symmetry for experiments ANTIG and ANTIC-1 to ANTIC-4.
3/2
(a) For the flux ΠÌ and Π`ν via the generalized skewness ΠÌ /Π`ν (black symbols) and
std ratio σÌ /(σ`ν )3/2 (green symbols). The dashed line is `2 . (b) For the wavelet velocity
increments δW at different scales, using ratio of standard deviation to average for δW
(green symbols) and δW C (yellow symbols). For self-similar variables, the ratio should be
constant (dotted line), which is true for δW C but not for δW. The dashed line is `−0.095 .
4.2. Rescaling symmetry breaking and generalized skewness

3/2
A basic ingredient of K41 is that the skewness S = h(δu)3 i/h(δu)2 i is constant
3/2
or equivalently, the generalized skewness S1 ≡ ΠÌ /(Π`ν ) scales like S1 ∼ `2 . This
is very well satisfied by the experimental data up to at least the Kolmogorov scale,
as shown in figure 8(a). Moreover, the scaling law is also satisfied by the standard
deviations σÌ /(σ`ν )3/2 ∼ `2 . Because of the universal properties of the PDFs of Π`ν
3/2
and ΠÌ (3.7), this means that any generalized skewness Sq ≡ (ΠÌ )q /(Π`ν )q scales
like `2q .
On the other hand, if we apply the h rescaling symmetry (iv) to ΠÌ and Π`ν , we
notice that they are respectively rescaled by a factor λ3h−1 and λ2h−2 , so that all the
generalized skewness values Sq are rescaled by a factor λ2q , in agreement with their
observed scaling. The minimal hypothesis to explain the experimental behaviour of
the high-order properties of the energy cascade is therefore that it is statistically
invariant by h rescaling, with no special emphasis on a particular h so far.
Furthermore, if one compares the individual behaviours of ΠÌ and Π`ν under h
rescaling symmetry (iv) and compare these with the scaling behaviours of the K41
theory, we see that they are compatible with respectively 3h − 1 = 0 and 2h − 2 =
−4/3, i.e. h = 1/3 for both. This mean that the rescaling symmetry (iii) is broken, so
that viscosity becomes irrelevant (in agreement with the ‘inertial range’ assumption)
and only h rescaling (v) with h = 1/3 should be considered. There is however no
understanding at this stage of how the special value h = 1/3 is selected with respect
867 P1-14
Beyond K41
(a) (b)
5 10-2
P(|ø|´-1/3˙2/3)
4
|ø|´-1/3˙2/3
3
2
10-4
1
0
50
50
Y/ 0 0
˙ 10-6 0
˙
-50 X/ 2 4 6 8
|ø|´-1/3˙2/3
F IGURE 9. Non-dimensional vorticity magnitude (a) and PDF of non-dimensional vorticity

magnitude (b) in the experiment CONC-3D. The figures are drawn using unpublished data
collected by the EXPLOIT collaboration (F. Daviaud, B. Dubrulle, P. Debue, V. Valori,
J.-P. Laval, J.-M. Foucaut, Ch. Cuvier, Y. Ostovan), with permission.
to other possible values. For this, we need to dive more deeply into the energy
cascade and its properties.
4.3. Energy cascade and inhomogeneity

Taking into account the K41 scaling, we can take the limit ` → 0 of the KHM
equation (3.1) to get
1
∂ hu2 (x, t)i −
2 t
hf · ui = −νh(∇u)2 (x, t)i, (4.1)
which is nothing but the global energy budget equation (2.2). Small-scale behaviour
of the energy cascade picture has, therefore, an immediate consequence on the
enstrophy scaling. Indeed, in homogeneous turbulence, the enstrophy hω2 (x, t)i =
h(∇u)2 (x, t)i, so we get from (4.1)

Ω ≡ hω2 (x, t)i = . (4.2)
ν
For this equality to hold, it means that there is in the flow at least one point where
√
the magnitude of the vorticity ω = ∇ × u exceeds the threshold: |ω| > /ν =
1/3 η−2/3 . There are in fact many points where this happens, as can be seen in
figure 9(a), where peaks of vorticity, over regions of size 30η, exist. Overall, the
vorticity field is very intermittent, meaning that if the homogeneity postulated by
Kolmogorov holds, it is only valid in a statistical sense (Frisch 1996).
4.4. Vortex stretching and 1/3 rescaling

Another measure of the intermittency of the vorticity field is provided by the PDF
of the magnitude of the vorticity, shown in figure 9(b). It is highly non-Gaussian,
867 P1-15
B. Dubrulle
with tails extending over values much larger than 1/3 η−2/3 . Such behaviour requires
a mechanism of amplification of vorticity: indeed, the forcing injects a typical value
of the velocity (L)1/3 , resulting in a typical vorticity of 1/3 L−2/3 , smaller than the
threshold by a value (η/L)2/3 . The mechanism has been identified by Taylor (1938).
It is based on the observation that, in the inviscid limit, both vorticity ωi and lines
ì are dynamically stretched by the velocity field, according to
∂t Ai + uj ∂j Ai = Aj ∂j ui , (4.3)
where Ai = ωi or ì . In a first approximation, let us ignore viscosity and concentrate
on the Lagrangian kinematic of vorticity in the frame where the tensor ∂j ui is
diagonal ∂j ui = Diag(λ1 , λ2 , λ3 ), with |λ1 | > |λ2 | > |λ3 |. Due to incompressibility,
λ1 + λ2 + λ3 = 0, so that either all the λi are real, and λ1 λ3 6 0, or one is real,
equal say to λ1 , and the other two are complex conjugates, equal to λ2 = λ∗3 , with
λ1 × real(λ2 ) 6 0. In such a frame, the kinematic behaviour associated with (4.3) is
simply
Dt Ai = Diag(λ1 A1 , λ2 A2 , λ3 A3 ). (4.4)
A blob of initial vorticity (ω1 , ω2 , ω3 ) with characteristic dimension (δ1 , δ2 , δ3 )
will experience two processes: its magnitude along the direction corresponding
to the eigenvalue with largest positive real part (say λ1 ) will be exponentially
amplified ω1 = exp(λ1 t), while its dimension along the direction corresponding
to the eigenvalue with largest negative real part (say λ3 ) will be exponentially
compressed δ3 = exp(−|λ3 |t). During such a process, the maximum vorticity ωmax
will then scale like
ωmax = δ3hω , (4.5)
with hω = −λ1 /|λ3 |. Direct numerical simulations of this inviscid kinematic stage
are difficult, since they require tracking of exponentially decreasing scales. A clear
observation of this amplification process in a vorticity pancake was achieved recently,
using combined high resolution direct numerical simulations (Agafontsev, Kuznetsov
& Mailybaev 2016), and vortex line dynamics (Agafontsev, Kuznetsov & Mailybaev
2017). They observe the scaling of (4.5), with hω = −λ1 /|λ3 ≈ −0.66. This kinematic
amplification cannot hold indefinitely, since there are at least two limiting processes:
viscosity and self-stretching. They arise as soon as either the smallest scale reaches
the Kolmogorov scale or when the vorticity produces a local self-stretching that can
oppose the velocity field stretching. This produces for example a rotation of vorticity
field that becomes aligned with the eigenvalue of intermediate magnitude (Pumir &
Siggia 1992), thereby saturating the vorticity amplification.
If we assume that the maximum vorticity is attained when the smallest scale
reaches δ3 ∼ η, the Kolmogorov scale, we observe that the vorticity amplification
mechanism discovered by Agafontsev et al. (2017) results in ωmax ∼ η−2/3 ,
i.e. precisely the scaling required by the Kolmogorov cascade picture and corresponding
867 P1-16
Beyond K41
to a 1/3 rescaling symmetry for the velocity field. Such mechanism requires that
the eigenvalues of the tensor ∂j ui are in a ratio [2 : 1 : −3]. This ratio has indeed
been reported to occur frequently in turbulent flows, but other frequent ratios, such
as [3 : 1 : −4] have also been reported (da Silva & Pereira 2008), which would
correspond to hω = −3/4 (1/4 rescaling symmetry for the velocity field) if the
same kind of mechanism holds. This would mean that h = 1/3 is not the only
special exponent in a turbulent flow, and this would explain why we see other
exponents appearing in the scaling behaviour of the standard deviations of the
cascade components ΠÌ and Π`ν (figure 7b).
Are there more indications of the existence of a multiplicity of relevant scaling
exponents h? To answer this question, we need to explore local scaling properties
of the velocity field.
4.5. Global-scale symmetry breaking and intermittency

As pointed out by Muzy, Bacry & Arneodo (1991), a natural tool to characterize
local scaling properties of the velocity field is the wavelet transform of the tensor
∂j ui , defined as
Z
Gij (x, r) = dr ∇ j Φ` (r)ui (x + r), (4.6)
where Φ` (x) = `−3 Φ(x/`) is a smooth function, non-negative with unit integral.
For all applications in the sequel, we choose a Gaussian function Ψ (x) =
exp(−x2 /2`2 )/N where N is the normalization constant such that d x Ψ (x) = 1.
R
We then compute the wavelet velocity increments as

δW(u)(x, `) = ` max |Gij (x, r)|, ` = |r|. (4.7)
ij
As shown in Muzy et al. (1991), if the velocity field satisfies locally the h rescaling
symmetry, then δW(u)(x, `) ∼ `h . Due to the observed inhomogeneity of the
velocity field derivatives (see § 4.3), the direct validation of this property via local
fit of δW(u)(x, `) is not possible in turbulent fields. Instead, one can look for
statistical signatures of the h rescaling symmetry via scaling properties of the PDF
of δW(u)(x, `). As shown in figure 10(a), this PDF is non-Gaussian, with fat tails
that vary with the scale `. Both their mean and their standard deviation follow a
`1/3 scaling (figure 11a) with tiny deviations. This is suggestive of a 1/3 rescaling
symmetry. A closer inspection of the ratio of the std to the mean (figure 8b) shows
that it is however not constant, suggesting a breaking of the 1/3 rescaling symmetry.
This is confirmed by a plot of centred and reduced PDFs of δW(u)(x, `), that do
not collapse on a single curve, as evidenced in figure 10(a1): the tails become
wider as the scale decreases, showing that some small-scale process is responsible
for the breaking of the 1/3 rescaling symmetry.
A quantitative measure of this effect is provided by the structure functions
Sp = h(δW)p i. If the 1/3 rescaling symmetry holds exactly, one expects them to be
867 P1-17
B. Dubrulle
(a) (a1) log10(¶/˙) (b) (b1)

1 3.0 1
1 1
0 0
log10(P)
log10(P)
-1
2.5 -1
-2
0 0 -2
-3
2.0 -3
-5 0 5 10
-5 0 5 10 15
log10(P)
∂W(u)(x, ¶)* ∂W(u)(x, ¶)C*

-1
1.5 -1
-2 1.0
-2
0.5
-3
-3
0 0.2 0.4 0.6 0 0.2 0.4 0.6
∂W(u)(x, ¶) ∂W(u)(x, ¶)C
F IGURE 10. PDFs of wavelet velocity increments at different scales, coded by colour, for
experiments ANTIG and ANTIC-1 to ANTIC-4 (see figure 11 for corresponding mean
and standard deviation and table 1 for symbols). (a) PDF of δW. Inset (a1): centred and
reduced PDF of δW at different scales, i.e. PDF of δW∗ = (δW − hδWi)/std(δW). (b) PDF
of δW C corresponding to δW conditioned on locations of low values of |D1xI
|. Inset (b1):
reduced and centred PDF of δW . C
(a) (b)
101
¶1/3 ¶1/3
Std(∂W), std(∂W C)
¯∂W˘, ¯∂W C˘
¶1/3
¶1/3 ¶
¶ 100
100
100 102 100 102

¶/˙ ¶/˙
F IGURE 11. Scale variation of the mean and standard deviations of the PDFs of wavelet
velocity increments for experiments ANTIG and ANTIC-1 to ANTIC-4. (a) Average:
hδWi: green symbols; hδW C i: yellow symbols. (b) Standard deviation std(δW): green
symbols; std(δW C ): yellow symbols.
power laws, Sp ∼ `ζ ( p) , with ζ (p) = p/3. These quantities are plotted in figure 12(a),
for p = 1 to 6. One sees that they indeed follow a power law with an exponent that
is displayed in figure 13(a). It is increasing with p and deviates further from the
straight line p/3 as p increases. Note that we find ζ (3) = 0.8 and not ζ (3) = 1 that
would naively have been expected from K41. Indeed, the prediction of K41 is only
for the scaling of the velocity increment of order 3. Here, ζ (3) corresponds to the
867 P1-18
Beyond K41
(a) (b)
1015
1010
Sp/´p/3˙Ω(p)
SpC/´p/3˙Ω (p)
C
1010
105
105
100 101 102 103 100 101 102 103

¶/˙ ¶/˙
F IGURE 12. Scale variation of the non-dimensional wavelet structure function of order
p = 1 to p = 6 for experiments ANTIG and ANTIC-1 to ANTIC-4. The structure functions
have been shifted by arbitrary factors for clarity and are coded by colour: p = 1: magenta
symbols; p = 2: green symbols; p = 3: light blue symbols; p = 4: red symbols; p = 5: blue
symbols; p = 6: orange symbols. (a) Structure functions for δW. (b) Structure functions for
δW C corresponding to δW conditioned on locations of low values of local energy transfer
at the resolution scale |D1x
I
|. The dashed lines are power laws with exponents shown in
figure 13(a). Adapted from Debue et al. (2018).
scaling of the absolute value of the wavelet increment. There is no prediction for
this quantity in K41. However, we can note that if we consider a rescaled scaling
exponent, ζr (p) = ζ (p)/ζ (3), (this is the exponent we may get by looking at the
scaling of Sp versus S3 , a technique called extended self-similarity (Benzi et al.
1993)) we guarantee that ζr (3) = 1, and we obtain something that appears to be
more universal (Arneodo et al. 1996). The deviation from the Kolmogorov law of
the scaling exponent, ξ V = ζ (p) − p/3, is shown in figure 17(b). It is well fitted by
a parabola.
Such a shape is puzzling, and implies strange pathologies for values of δW that
deviates strongly from the mean, as noted in Frisch (1996). A way to study
the behaviour of the distributions of δW is through the function U(p, `) =
h(δW)p+1 i/h(δW)p i. Indeed, as p increases, this function samples values of δW
that deviate increasingly from the mean. Taking into account the observed
scaling properties of δW, we get the scaling U(p, `) ∼ `1( p) with 1(p) =
1/3 + ξ V (p + 1) − ξ V (p) which becomes negative as soon as p > p∗ ≈ 9. This
means that if we take first the limit ν → 0 and then ` → 0, the corresponding
values U(p > p∗ , `) can take arbitrary large values. This is another indication that
the process that is responsible for the breaking of the global symmetry that occurs
at small scale, and involves very large values of the velocity gradient, in the limit
ν → 0. Where does this phenomenon come from?
867 P1-19
B. Dubrulle
(a) 3.0 (b)

0
2.5
-0.5
≈˜( p/3), ≈ I( p/3)

2.0
Ω( p)
1.5
-1.0
1.0
-1.5
0.5
-2.0
0 2 4 6 8 10 0 2 4 6 8 10
p p
F IGURE 13. Scaling exponents as a function of order for experiments ANTIG and
ANTIC-1 to ANTIC-4. (a) Scaling exponents ζ (p) of the wavelet structure functions
of δW (green symbols) and δW C (yellow symbols). The exponents have been computed
through a least square algorithm upon ζ (p), minimizing the scatter of the non-dimensional
structure functions of different experiments with respect to the line (`/η)ζ ( p) . The green
dotted line is the function minh (hp + C(h)) with C(h) given in figure 27(a). The black
triangles are the rescaled scaling exponents ζr (p) = ζ (p)/ζ (3). (b) Scaling exponents
of the structure function of |DII | and of |DIν |: ξ I (p/3): red circle; ξ ν (p/3): blue circle.
The red dotted (respectively dashed) lines are log-normal fits ξ I (p) = 9bp(1 − p)/2 with
b = 0.045 (respectively 0.065). The blue dotted (respectively dashed) lines are log-normal
fits ξ ν (p) = 2bp − 2bp2 with b = 0.045 (respectively 0.065). For comparison, we also
reported other scaling exponents: ξ V = ζ (p) − p/3 (green squares); τrsh = ζ (p) − ζ (3)p/3
(black triangles); τ (p/3) for DNS (black stars) and experiment ANTIC-4 (magenta stars),
computed from figure 17. The green dot-dashed line is a parabolic fit: ξ V = −0.015p2 −
0.046p + 0.06. Adapted from Debue et al. (2018).
4.6. Time-reversal breaking and enstrophy blow up

Let us come back to the enstrophy balance (4.2) – a by-product of the cascade
picture – and explore its consequences as we let ν → 0. For this, we need to
characterize the behaviour of , the rate of injected power, as a function of the
viscosity ν. In the von Kármán flow (figure 1b), where the fluid is forced by rotating
paddles, this has been done recently using measurements in experiments with the
same geometry, but different fluids and tank sizes (Saint-Michel et al. 2014). It
was found that depends on the fluid viscosity ν and on the geometry, size L and
rotation velocity Ω of the paddle. However, a remarkable universality appears in the
limit of large Reynolds numbers, Re = L2 Ω/ν, where the flow becomes turbulent.
In this limit, the energy injection rate becomes independent of viscosity, and scales
like
lim = 2Kp L2 Ω 3 , (4.8)
Re→∞
where Kp is a parameter that only depends on the geometry of the paddles for a
given the topology of the mean flow (figure 14). This universality carries on to
867 P1-20
Beyond K41
(b)
(-)b
1
z/R 0
-1
Kp (-)s
(s)
1
-1
10
z/R 0
(+)s
-1
x/R
102 104 106 108

Re
F IGURE 14. Non-dimensional energy injection rate /2L2 Ω 3 as a function of Reynolds

number Re in von Kármán flow using different fluids and different forcings in different
experiments. For Re > 104 , the mean flow is bistable for the (−) forcing, and displays
two different topologies: (i) a symmetric one, noted (s), shown on the right-south of the
main figure; (ii) a non-symmetric one, noted (b) shown on the right-north of the main
figure. For a given (−) forcing, the energy injection rate is much higher if the flow is in
the (b) topology than in the (s) topology. The colour codes the various forcings and flow
topologies: red symbols: (+) forcing according to figure 1(c), with symmetric mean flow
topology; blue and green symbols: (−) forcing in respectively symmetric and bifurcated
mean flow topology. The symbols code the fluid: pentagons: mixture of water and glycerol,
at different concentrations; square: water; diamonds: liquid sodium; left-pointing arrows:
helium 4 at T = 2.3K. Right-pointing arrows: helium 4 at T = 2 K, i.e. in its superfluid
phase. The grey symbols are measurements performed in mixture of water and glycerol,
with impellers similar to the one shown in figure 1(c), except that they are fitted with
16 blades, instead of 8. The diamonds are unpublished data from the VKS experiment.
Courtesy VKS collaboration (M. Faure, N. Bonnefoy, S. Miralles, N. Plihon, J.-F. Pinton,
Ph. Odier, G. Verhille, M. Bourgoin, S. Fauve, F. Petrelis, M. Berhanu, N. Mordant,
B. Gallet, S. Aumaitre, F. Daviaud, A. Chiffaudel, R. Monchaux, P. Gutierrez). The arrows
are obtained from data from the SHREK experiment, with special help from B. Rousset.
Adapted from Saint-Michel PhD thesis (Saint-Michel 2013).
the Re = ∞ limit as proved by recent experiments using super-fluid helium 4 at

low temperature (Saint-Michel et al. 2014). The behaviour described in (4.8) is
well known in the turbulence community and is one of the most robust laws in
turbulence. It is sometimes referred to as the zeroth law of turbulence. It is however
a rather puzzling law: in the cascade picture, all injected energy is dissipated into
heat, i.e. into entropy. Production of entropy is natural in a dissipative system, and
is the signature of the breaking of the time-reversal symmetry – in the language of
statistical physics, we would call it an order parameter of the time-reversal symmetry.
867 P1-21
B. Dubrulle
As we have seen in § 4.1, this symmetry is explicitly broken by viscosity. Then,
the zeroth law of turbulence means that in a turbulent fluid, at large Reynolds
numbers, the order parameter becomes independent of the process that breaks the
symmetry. In statistical physics, this behaviour is referred to as a spontaneous
symmetry breaking, while in modern field theory language, it is called an anomaly
(Falkovich, Gawȩdzki & Vergassola 2001), so that the constancy of /L2 Ω 3 is
called the dissipation anomaly.
This anomaly has important physical implications in the fluid. When plugged
into the enstrophy balance (4.2), it leads to limν→0 Ω = ∞ i.e. enstrophy blow
up. This means that there is at least one location in the flow where the vorticity
becomes infinite, i.e. that there is a singularity in the flow in the inviscid limit. This
conclusion may be disturbing for physicists, but not so much for mathematicians,
who have been hunting singularities in the Navier–Stokes equation and its inviscid
limit, the Euler equation, for almost a century, starting from Leray (1934). In
any case, it is an indication that suitable mathematical tools should be used to
understand in greater details the physics of turbulence. This will be the topic of
§ 6.
Before that, we note that the enstrophy blow up in the inviscid limit is a very
strong argument against the universality of small-scale statistics K41 theory, as noted
by Landau. Indeed, K41 assumes that everything only depends on the viscosity ν
and on , the mean energy dissipation rate. The enstrophy blow up in the inviscid
limit means that the energy dissipation is subject to large statistical fluctuations
that are not necessarily universal, as can be directly observed from figure 6(b1)
where the distribution of the energy dissipation displays wide tails. This means
that the fluctuations of the energy dissipation cannot be ignored and may break the
universality of K41. This remark led Obukhov and Kolmogorov (1962) to build
a ‘refined’ theory (hereafter named K62) that takes explicitly into account these
fluctuations.
5. The Kolmogorov–Obukhov refined similarity hypothesis

5.1. Summary of the theory K62
Kolmogorov considers the quantity ` , defined as the viscous dissipation ν averaged
over a ball of size `, and assumes further that ` has a log-normal distribution, with
mean and variance of log given by
` = , ∀`, η 6 ` 6 L, (5.1)
σ (ln(` )) = −2A − 2µK62 ln(`/L),
2
(5.2)
where A is a non-universal constant that depends on the flow geometry, and µK62 is a
universal constant that characterizes the fluctuations. Since the characteristic function
of ln ` gives the moments of and is quadratic for a Gaussian, it is straightforward
867 P1-22
Beyond K41
to show that the two hypotheses result in a universal scaling for all moments of `
according to
µK62 p(1−p)
`
` ∼ p ep(1−p)A .
p
(5.3)
L
From this, we can deduce the scaling law of two other interesting quantities:
" #
` −2µK62
σ (` ) ≡ `2 − ` = e
2 2 2 −2A
−1 , (5.4)
L
ln(` ) ≡ ∂p (` )|p=0 = ln() + µK62 ln(`/L) + A.
p
(5.5)
Kolmogorov then postulates local universality, so that locally, all quantities
only depends on `, ν and ` . From this, he can define the local velocity
scales as U` = `1/3 `1/3 , dissipative scale η` = ν 3/4 `−1/4 and Reynolds number
Re` = Ù` /ν = (`/η` )4/3 . He then obtains
(δu` )p = (`` )p/3
µK62 (1−p/3)p/3
`
= Cp (`) p/3
. (5.6)
L
The scaling of (5.6) means that the global self-similarity is broken, since the rescaled
moments (δu` )p /((δu` )3 )p/3 scale like `τK62 ( p) with τK62 (p) = µK62 (1 − p/3)p/3.
Moreover, it provides the origin of the symmetry breaking, as resulting from
large fluctuations of the local energy dissipation rate over a ball of size `. In
such a process, a new length scale (namely L) now appears to play a major role,
contributing to scale symmetry breaking.
5.2. Test of refined similarity hypothesis

The refined similarity hypothesis can be tested easily on numerical data. The test
with experimental data is more tricky, as it requires measurements of 3 components
of velocities at a resolution at least equal to the Kolmogorov scale. Our SPIV data
in glycerol are resolved down to the Kolmogorov scale, but they are not rigorously
three dimensional, since they are measured only in a plane. They can however be
used to compute a pseudo-energy dissipation as ν = −ν(∂P u)2 , where ∂P means the
derivative taken only onto the plane of measurements. Given the local instantaneous
energy dissipation ν (x, t), an estimate of ` is through a convolution through the
smoothing function we used for wavelets
Z
` (x, t) = dr Φ` (r)ν (x + r, t), (5.7)
ν (x, t) = ν (∂u)2 (x, t). (5.8)

Observe that for ` → 0, ` → ν and that for ` → L (L being the size of the volume
containing the fluid), ` → Pdiss /ρL3 , the instantaneous energy dissipation rate.
867 P1-23
B. Dubrulle
(a) ln(¶/L) (b) 1

0
-3.5 0
-2 -1
-4.0
-2
-4
log10(P)
-4.5
-3
-6 -5.0 -4
-5
-8 -5.5
-6
-10 -6.0 -7
50 100 150 200 250 300 -5 0 5
´¶ (ln(´¶) - ¯ln(´¶)˘)/ß(ln(´¶))
F IGURE 15. Probability distribution functions of the pseudo-energy dissipation averaged

over a ball of size ` in experiment ANTIC-4 at various scales `/L, coded by colour
(see colour bar). The magenta squares are for the non-averaged pseudo-dissipation ν .
The yellow symbols, for the time averaged pseudo-dissipation Pdiss /ρL3 . (a) PDF of ` .
(b) Centred reduced PDF of ln(` ). The black dashed line is a log-normal with zero mean
and unit variance. The grey crosses are centred and reduced distributions of ` in DNS
A, at scales corresponding to `/L = 0 and `/L = nη/L, with n = 1, 3 and 10.
The PDF of the resulting distributions, for various `, is provided in figure 15(a),
along with the distributions of ν and Pdiss /ρL3 . One sees that they vary widely
in shape, as ` is varied, with a variance decreasing as scale is increased. To test
the hypothesis of K62, we compute the PDF of ln(` ) for various `/L, shown in
figure 15(b). They are indeed close to parabolic, with deviations that could be due
noise or projection effect. By comparison, the PDF of ln(` ) for the numerical
simulation DNS A is indeed perfectly parabolic, for all scales. To investigate the
accuracy of K62 to describe the data, we compute the mean and variances of ` and
ln ` and plot them as a function of `/L in figure 16(a,b), for both numerical and
experimental data. The average of ` is indeed constant (equation (5.1)), while its
variance decreases and is well described by (5.4) using µ = 0.13 and A = 0.36. With
these two numbers, we can then fit the mean and variances of ln(` ) using (5.5)
and (5.2) without adjusting parameters. We see that the fit is fair for experiments,
but rather good for the numerical data.
Consistency with K62 can be further checked by computing moments of higher
order for ` as a function of `. They are reported in figure 17(a). In the inertial range
ln(`/η) > 4, they follow the power law h(` )p/3 i ∼ p/3 `τ ( p/3) , with τ (p/3) reported
in figure 17(b) for both experiments and DNS. They are parabolic, as suggested by
K62, and well fitted by the exact log-normal shape τ (p/3) = µK62 (1 − p/3)p/3, with
µK62 = 0.13. The moments h(` )p/3 i can also be well fitted without any adjustable
parameter by (5.3), with µK62 and A computed using the variance of ln(` ). From
867 P1-24
Beyond K41
(a) 1.5 (b) 2
1.0 1
ln¯´¶/´˘, ln (ß 2(´¶/´))
¯ln(´¶/´)˘, ß 2(ln(´¶))
0.5 0
0 -1
-0.5 -2
-1.0 -3
-8 -6 -4 -2 -8 -6 -4 -2
ln(¶/L) ln(¶/L)
F IGURE 16. Mean (red symbols) and variances (green symbols) of PDF of (a) ` and (b)
ln(` ) in experiment ANTIC-4 as a function of scale `/L: circle: ν ; stars: ` ; triangles:
Pdiss /ρL3 . In (a), the dashed line is a fit using the shape given by (5.4), with µK62 = 0.13
and A = 0.36. These two numbers are then used to compute the dotted (respectively the
dashed line) in (b), via (5.5) (respectively (5.2)). The coloured full lines are obtained using
DNS A.
all this, we see that the log-normal hypothesis formulated by Kolmogorov is fairly
well satisfied by numerical and experimental data.
However, when we compare τ (p/3) and ξ V = ζ (p) − p/3 computed from the
wavelet analysis in § 4.5, we observe a rather large discrepancy, in contradiction with
the Kolmogorov refined similarity hypothesis (5.6). We can argue that the problem
is due to the fact that ζ (3) is not equal to one (see § 4.5), meaning that ξ V (1) is not
zero, as necessary in the refined scaling hypothesis. We may correct that by using
τrsh (p/3) = ζ (p) − ζ (3)p/3, which is indeed closer to τ (p/3), but there is still a
discrepancy. Where does the problem come from?
5.3. Problems and issues set by K62

Let us come back to the mathematics behind K62. In the Kolmogorov cascade
picture, the energy is transferred locally from a scale ` to a scale Γ ` (Γ 6 1) in
a self-similar way, so that the distribution of ` /Γ ` only depends on the ratio of
scales Γ . Writing L /Γ N L = (L /Γ L ) · · · (Γ N−1 L /Γ N L ), we see that for any ` = Γ N L,
the quantity ln(` /L ) is the sum of N = ln(L/`)/ ln(Γ ) variables with identical
distributions
ln(L/`)/ ln(Γ )
ln(` /L ) = Σn=1 ln(Γ n L /Γ n−1 L ). (5.9)
When N = ln(L/`)/ ln(Γ ) → ∞, and if the distributions are independent, we can
then use large deviation theory (Kramer’s theorem) (Touchette 2009) to write the
probability of observing ln(` /L ) as
P(ln(` /L ) = x) ∼ e−NF(x/N) , (5.10)
867 P1-25
B. Dubrulle
(a) 5 (b) 1.0
†´(p/3), †(p/3),†rsh(p/3)
4 0.5
ln(¯(´¶/´)p/3˘)
3 0
2 -0.5
1 -1.0
0 -1.5
-1 -2.0
-8 -7 -6 -5 -4 -3 2 4 6 8
ln(¶/L) p
F IGURE 17. Test of refined similarity hypothesis. (a) Structure functions of ` as a

function of `/L in experiment ANTIC-4 (symbols): `/L: circle: ν ; stars: ` ; triangles:
Pdiss /ρL3 , and in DNS A (continuous line). The structure functions are coded by colour:
p = 1: blue; p = 2: orange; p = 3: yellow; p = 4: magenta; p = 5: green; p = 6: light
blue; p = 7: red; p = 8: blue; p = 9: orange. The coloured dashed lines are (5.3), with
µK62 and A computed using variance of ln(` ). (b) Scaling exponents of the structure
functions: τ (p/3) for experiment ANTIC-4 (magenta stars); for DNS A (black stars).
The black dashed line is µK62 (1 − p/3)p/3, with µK62 = 0.13, obtained in figure 16.
These values do not coincide with the quantity τ (p/3) = ζ (p) − p/3 (green squares),
and τrsh (p/3) = ζ (p) − ζ (3)p/3 (black triangles) inferred from figure 13(a) using directly
wavelet velocity increments, that are best fitted (green dotted line) by µ(1 − p/3)p/3 with
µ = 0.24 > µK62 , in contradiction with K62.
where F is the large deviation function that only depends on the distribution of
` /Γ ` . The highest probability (the most likely event) is obtained for the value x
that is connected to the z∗ , the location of the maximum value of F, via x = Nz∗ .
For values of x/N close to z∗ , the function F is parabolic, so that F(z) = F(z∗ ) +
F 00 (z∗ )(z − z∗ )2 /2 and
∗ )−(F 00 (z∗ )/N)(x−Nz∗ )2 /2
P (ln(` /L ) = x ≈ z∗ ln(L/`)) ∼ e−NF(z , (5.11)
corresponding to a log-normal distribution for /L , with a variance ln(L/`)/F 00 (z∗ ).
This is K62 theory, with µK62 ∝ 1/F 00 (z∗ ). However, for values far from the most
probable value, the large deviation function F can differ from a parabola, and
deviation from log-normal can be observed.
This argument is only valid provided the distributions of Γ p−1 ` /Γ p ` for different
p are independent. If they are correlated, no general prediction about their sum can
be given (see Clusel & Bertin (2008) for solvable examples), and one can obtain
shapes different from log-normal. We thus see that the K62 theory is based on an
idealized treatment of the fluctuations that does not necessarily reflect reality.
Another questionable assumption holds in the interpretation of (5.6), that allows us
to connect the velocity structure function to the quantity ` . Indeed, in the original
867 P1-26
Beyond K41
K41 theory, ` was inferred from KHM equation as the constant scale-to-scale
energy transfer. This remark led Kraichnan (1974) to suggest that if such a relation
holds, ` should represent some well-defined local energy flux Π` rather than a
local energy dissipation averaged over a volume of linear dimension `. Since K62
is not derived directly from the Navier–Stokes equations, it is not obvious how
to define precisely the expression and properties of Π` . To make further progress,
we need to find a generalization of K41 that (i) provides and expression of such
local energy flux; (ii) does not suffer from the main drawbacks of K41, namely
that does not assumes homogeneity, nor stationarity, that allow for breaking of the
time symmetry and (iii) takes into account the fluctuations of energy dissipation
and allows for the breaking of the self-similarity. This generalization was derived
in 1990 by two French mathematicians, Duchon and Robert. The purpose of the
next section is to summarize their derivation, and explore its outcomes.
6. Weak formulation of Kármán–Howarth–Monin equation

6.1. Derivation of the weak Kármán–Howarth–Monin equation
6.1.1. Why a weak formulation?
In previous sections, we have identified many clues about a pathological
behaviour of the velocity field in turbulent flows: development of large gradients,
inhomogeneity and intermittency of the dissipation, spontaneous breaking of the
time-reversal symmetry . . .. Even the simple Kolmogorov picture involves lack of
regularity for the velocity field. Indeed, we have
|h(δu)3 i| 6 h|δu|3 i. (6.1)
Let us further assume that there exist an h and a C, such that ∀x and ` 6 `0
|δu| 6 C`h . (6.2)
In mathematical language, this means that the velocity field is uniform Hölder
continuous with exponent h. Substituting in (6.1), we then see that
|h(δu)3 i| 6 C3 `3h , (6.3)
so that Kolmogorov scaling cannot be satisfied if h > 1/3, because h(δu)3 i will tend
to zero always faster than `. This means that there is at least one point in the flow
where the inequality (6.2) with h > 1/3 is violated, possibly satisfied only with some
h 6 1/3. At such a point, the field is irregular, since velocity gradients can blow up
like `h−1 in the limit ` → 0. We call such fields ‘rough’, in opposition to ‘smooth’
fields, characterized by a Hölder continuous exponent h = 1 everywhere.
Rough fields are delicate to deal with, especially in the context of partial
differential equations such as the Navier–Stokes equations, because derivatives
are not necessarily well behaved. A suitable tool to deal with them was invented by
867 P1-27
B. Dubrulle
Leray (1934) and named ‘weak formulation’. The main idea is to make a detour via
the scale space, and work with a smoothed version of the initial field (a ‘mollified’
field), over a characteristic scale (resolution) `. At any given resolution `, the
mollified field is sufficiently regular, so that all classical tools and the manipulation
of the analysis of vector fields are valid. Limiting behaviours as the resolution
` → 0 can then be used to infer results and properties for the rough field.
6.1.2. How to implement a weak formulation?

The implementation of a weak formulation requires the introduction of a
smoothing operator (or test function) φ(x), with suitable properties: it must be
smooth (at least twice differentiable for Navier–Stokes) with R compact support
3
on R , even, non-negative, spatially localized and such that dr φ(r) = 1. The
smoothing at a resolution ` is then achieved through the function φ` , defined as
φ` (r) = `−3 φ (r/`). For any given rough field u(x, t), the mollified field is defined
as Z
uì (x, t) = dr φ` (r)ui (x + r, t). (6.4)
The properties of φ guarantee that when ` → 0, uì → ui . Moreover, the velocity

derivative of the rough field ui can be written as
Z
`
∂j ui (x, t) = − lim dr ∂j φ` (r)ui (x + r, t), (6.5)
`→0
as well as any higher-order derivative of ui , as long as φ is sufficiently regular.

The weak formalism has some interesting connections with classical tools of
turbulence. First, equation (6.4) shows that the mollifying process is a coarse
graining. As such, it averages out fine details about the fields while keeping
information about large scales. Formally, the mollified velocity can also be seen
as a continuous wavelet transform of the velocity u with respect to the wavelet
φ. Note, however, that since we have chosen φ to be of unit integral, it is not
admissible, meaning that the wavelet transform is not invertible.
6.1.3. Advantages of the weak formulation

The link between the weak formulation and wavelet transform unveils several
advantages of its application to turbulent rough fields. A wavelet transform is a
kind of ‘local Fourier transform’. As such, it is more suitable for application onto
non-homogeneous fields and will naturally deal with the observed breaking of the
space translation symmetry (§ 4.3). A wavelet transform is also by construction
‘locally self-similar’. It is therefore the relevant tool to measure local Hölder
exponents, as proved by Arneodo and his group (Muzy et al. 1991). Specifically, it
can be shown that if a field is locally Hölder continuous with exponent h around
x0 , then its wavelet transform will scale locally like |uì (x0 , t)| ∼ `h . (Rigorously, this
867 P1-28
Beyond K41
is valid only when the wavelet has enough vanishing moments, namely more than
the degree of a local polynomial corresponding to the regular part of the Taylor
expansion of u near x0 .) This means that the weak formulation will be able to keep
track of the global-scale symmetry and its violations (§ 4.5)).
I hope that the reader is now convinced that the weak formulation, and its
associated continuous wavelet transform, is the natural tool to generalize the KHM
equation into a ‘local’ version, that will help to go beyond K41 and K62. Let
us now summarize the main steps of Duchon and Robert’s derivation (Duchon &
Robert 2000).
6.1.4. Summary of Duchon–Robert derivation

We first apply the wavelet transform to the Navier–Stokes equations (1.1), to get
a classical coarse-grained equation
1 `
∂t uì + u`j ∂j uì = −∂j τij` − ∂i p + ν∂jj uì , (6.6)
ρ
where τij` = (ui uj )` − uì u`j is the classical Reynolds stress tensor. We multiply (1.1)
and (6.6) by u` and u, respectively, and add them together; after the rearrangement
of terms we obtain the following balance equation:
∂t ui uì + ∂i Ti = E ` + νui ∂ 2 uì + νuì ∂ 2 ui , (6.7)
where
Ti = uj u`j ui + (p` ui + puì )/ρ,
)
(6.8)
E ` = ui uj ∂i u`j − uj ∂i (ui uj )` .
Developing terms and using properties of φ, it is simple to check that E ` can be

split into
Z
2E = −∂i (ui uj uj ) + ∂i (ui (uj uj ) ) − dξ ∇φ ` (ξ ) · δξ u(δξ u)2 ,
` ` `
(6.9)
where δξ u is the velocity increment over a distance ξ and ∇ the gradient over ξ .
We can also split
uì ∂ 2 ui = ∂j (uì ∂j ui − ui ∂j uì ) + ui ∂ 2 uì . (6.10)
Finally, it is straightforward to check that
Z
1 1
dξ ∇ 2 φ ` (ξ )(δξ u)2 = ∂ 2 (ui ui )` − ui ∂ 2 uì . (6.11)
2 2
After the rearrangement of terms, we finally get a local balance equation for the
quantity E` (x) ≡ ui uì /2 as
∂t E` (x) + ∂j Jj = −DÌ − D`ν , (6.12)
867 P1-29
B. Dubrulle
where
Ji = ui E` + 12 (p` ui + puì ) + 14 ([ui uj uj ]` − ui [uj uj ]` ) − ν∂i (uì ∂j ui − ui ∂j uì ), (6.13)
Z
1
D` =
I
dξ ∇φ ` (ξ ) · δξ u(δξ u)2 , (6.14)
4
ν
Z
D`ν = dξ ∇ 2 φ ` (ξ )(δξ u)2 . (6.15)
2
We name such a balance equation the ‘weak KHM equation’, denoted WKHM. Its
physical interpretation is provided in § 6.2.2
6.2. Properties of WKHM

6.2.1. Robustness with respect to noise
WKHM is robust with respect to the addition of noise that is isotropic, Gaussian
and not correlated to the velocity – this is often true for velocity measurements, in
the absence of systematic errors. In such a case, the measured velocity increments
can be simply written as δumeas = δu + α, where δu is the true velocity increment
and α is the noise, such that for any (i, j, k) hαi i = hαi αj αk i = 0 and hαi αj i = N δij ,
where N is the noise amplitude. Since we further have
δumeas |δumeas |2 = δu|δu|2 + α|δu|2 + 2δu(α · δu) + 2α(α · δu) + δu|α|2 + δα|δα|2 ,
(6.16)
we get by statistical averaging
hδumeas |δumeas |2 i = hδu|δu|2 i + 3N hδui. (6.17)
If the velocity field is statistically homogeneous at the considered location, then
hδui = 0, so that all the noise contribution has been averaged out and there is no
noise amplification introduced by taking the divergence. In the same way, if the
noise has no spatial correlation, the statistical average guarantees that the noise
contribution is averaged out in h(δu)2 i, so that it can be differentiated twice without
noise amplification.
In the weak formulation, the gradient is not applied directly to the velocity
increments, but rather to the smooth test function, preceded by a local angle
averaging. The latter plays a role similar to statistical averaging for isotropic noise.
The convolution with the derivative of the smoothing function further guarantees
that there is no noise amplification when computing DI` and Dν` . This robustness
with respect to noise makes the quantities DI` and Dν` very interesting tools to
characterize local energy transfers in laboratory flows using e.g. particle image
velocimetry (PIV) measurements: it involves only velocity increments, which are
easily computed from the velocity field data obtained by a such technique. For
numerical flows, the potential gain is not to noise, but rather to the accuracy of
the description in regions where the velocity field presents a sharp discontinuity or
local strong variations, as illustrated in § 6.2.5.
867 P1-30
Beyond K41
6.2.2. Interpretation of WKHM as a local KHM
Equation (6.12) describes the evolution of E` (x), the point-split kinetic energy
at scale ` and at position x, through three main ingredients: (i) a spatial flux term
∇ · J, which describes how the input energy is transported within the flow; (ii) an
inter-scale flux DÌ , which describes how the energy cascades locally across the
length scales; and (iii) D`ν , which describes energy space transfer and dissipation
by viscosity. It is, in fact, a local non-random form of the classical KHM equation,
as already recognized in Duchon & Robert (2000). To see R that, we note that the
quantity E` (x) can also be written as E` (x) ≡ ui uì = (1/2) φ ` (ξ )ui (x)ui (x + ξ ) dξ ,
which may be interpreted as a local average over a ball of size ` of the quantity
e(x, r) = ui (x)ui (x + r), i.e. the local equivalent of the correlation function of KHM
equation, where the statistical average is replaced by a local average over scale. In
the same way, the inter-scale flux term DÌ can be written after integration by parts
as Z
1
DÌ = − dξ φ ` (ξ )∇ ξ · hδξ u(δξ u)2 i, (6.18)
4
so that it is the local equivalent of ΠÌ . Finally, we can perform two integrations by
parts on the viscous term D`ν to show that it is the local average over a ball of size `
of ν∇r2 e(x, r) so that it is the local equivalent of Π`ν . The spatial flux term ∇ · J has
no equivalent in KHM, because it describes a purely local effect: it vanishes when
integrated over space, if there are no input of energy at the flow boundary. However,
if turbulence is forced by boundaries, like in a plane Couette flow, its space average
will contribute to P` , the mean energy injection rate at scale `.
In the same way, we note that by space average, hE` (x)ix = hC(r)iB` , the average
over a ball of size ` of the correlation function. In the same way, the space averages
of DÌ and D`ν are exactly the averages over balls of size ` of ΠÌ and Π`ν . These are
the quantities that are plotted in figure 5 for both numerical and experimental data,
and that are shown to obey, on average over the space, the classical K41 scaling.
Being local in space and time, DÌ and D`ν are however natural candidates to trace
and understand the dynamics of energy cascades and dissipation in turbulent flows,
and their link with symmetry breaking and intermittency, as we see now.
6.2.3. Small-scale limit

WKHM allows us to define energy transfers and dissipation for any given scale
`. Before investigating their properties at various scales, let us focus on their limit
` → 0, which involves mathematical subtleties due to the possible roughness of the
velocity field. Consider for example the limit of u` . By construction, we have
Z Z
u (x) = dξ φ (ξ )u(x + ξ ) = u(x) + dξ φ ` (ξ )δξ u.
` `
(6.19)
At all points x where u is regular, we have δξ u = ξ · ∂u + O((ξ )2 ), so that for

sufficiently small `, u` (x) = u(x) + O(`). There is therefore no problem to take the
867 P1-31
B. Dubrulle
small-scale limit of u` , and we have lim`→0 u` (x) = u(x). When the velocity field
is rough, the limit is less trivial. At such points, we define the Hölder exponent
h < 1 (depending upon the considered point) such that the velocity field obeys the
condition
∃C ∀` 6 `0 , |δ` u(x)| 6 C`h . (6.20)
`
Substituting this into (6.19), we now see that for sufficiently small `, u (x) = u(x) +
O(`h ). So, as long as h > 0 – this includes the case h = 1/3 corresponding to K41 –
the small-scale limit of u` (x) is still simple and equal to u(x). Things become more
delicate for quantities involving derivatives, as they may now behave wildly in the
small-scale limit. We have indeed
Z Z
∂u` (x) = dξ φ ` (ξ )∂u(x + ξ ) = dξ φ ` (ξ )∇(u(x + ξ ) − u(x))
Z
= − dξ ∇φ ` (ξ )δξ u(x) = O(`h−1 ). (6.21)
Therefore, for h < 1, the derivative ∂u` (x) is unbounded and its limit (if it exists),
noted ∂u(x), may be infinite. Using (6.20), we then get
D`ν = O(ν`2h−2 ), (6.22)
DÌ = O(`3h−1
). (6.23)
The limit of the viscous term is
lim D`ν = ν , (6.24)
`→0
where ν = ν(∂u)2 is the local viscous dissipation (potentially infinite if ∂u is

infinite). We call D I = lim`→0 DÌ . From (6.23), we see that D I = 0 whenever
h > 1/3. For h 6 1/3, it is possible that D I is non-zero. What does it mean?
6.2.4. Inertial dissipation and Onsager’s conjecture

Consider WKHM and perform first the limit ν → 0, and then ` → 0. From
previous estimates (6.19), (6.21) and (6.23), we then get
∂ u2 + ∂ 12 u3 + pu = −D I .
1

2 t
(6.25)
This local energy budget shows that, even in the absence of viscosity, the energy is
not conserved and can be changed by the term D I , emerging from the roughness
of the velocity field. If D I > 0, we can rephrase this observation into Onsager’s
own words (Onsager 1949) as: in three dimensions a mechanism for complete
dissipation of all kinetic energy, even without the aid of viscosity, is available. Lars
Onsager was the first to make the connection between the regularity properties of
the velocity field and kinetic energy conservation in the Euler equations. His finding
remained unnoticed for a long time (Eyink & Sreenivasan 2006). It was not until
867 P1-32
Beyond K41
2000 that Duchon and Robert brought the attention of the turbulence community
to this mechanism and named it ‘inertial dissipation’. They proved that in two
dimensions, D I = 0. In three dimensions, the condition of existence of a suitable
weak solution only implies that the spatial average hD I i > 0. Note that, because
of (6.23), h = 1/3, the Kolmogorov value, is the maximum regularity condition
compatible with a non-zero inertial dissipation. This provides another indication of
why this value is so special in turbulence.
6.2.5. Illustration: inertial dissipation by a shock

A simple illustration of the concept of inertial dissipation can be given using
a a shock, the singularity appearing in the inviscid Burgers equation, sometimes
regarding as the ‘one-dimensional equivalent of Navier–Stokes equations’ (see Bec
& Khanin (2007) for a complete review). Its velocity field is defined in the interval
[−L, L] and is obtained as the limit of an exact solution of the viscous Burgers
equation via us (x, t) = limν→0 uν (x, t) = limν→0 (x − L tanh(Lx/2νt))/t, resulting in the
‘Khokhlov sawtooth solution’
us (x, t) = (x + L)/t − L 6 x < 0, (6.26)
us (x, t) = (x − L)/t 0 < x 6 L. (6.27)
The velocity field is singular at x = 0, where it exhibits a jump (a shock) of size
1u = 2L/t (see figure 18a). The velocity increments δ` u are proportional to `/t
whenever 0 is outside the range [x, x + `] and of the order 1u + O(`/t) whenever
0 is inside this range. Formally, the shock is therefore a singularity with Hölder
exponent h = 0. Plugging this into (6.14), we see that DÌ = O(`2 ) for x outside the
range [−`, `] and DÌ ∝ (1u)3 /L + O(`2 ) for x ∈ [−`, `], so that D I ∝ (1u)3 /L > 0:
the shock dissipates energy. Exact calculations (see e.g. Eyink (2007–2008)) show
that D I = (1u)3 /12L = , where = limν→0 ν(∂uν )2 , so that the discontinuity
dissipates exactly the same energy as the viscous solution. Figure 18(b) shows the
space/scale diagram of the local energy transfer DÌ for the shock solution. One
sees that, at a given scale, it is concentrated in the range [−`, `] around the shock.
As ` decreases, DÌ concentrates more and more around the shock, ‘pointing’ at
the singularity. If we now look at the behaviour of DÌ (0) with ` – the amplitude
of the local energy transfer at the singularity point – we see that it diverges like
1/`, i.e. DÌ (0) ∼ `3h−1 with h = 0 (figure 19a). Its space average converges towards
like hDÌ i − ∼ ` (figure 19b), which corresponds to the probability of hitting
a point-like singularity (co-dimension 1). By looking at the behaviour of DÌ with
decreasing scale, one sees that it is possible to infer where there is a singularity,
what is its Hölder exponent and its probability?
6.2.6. Effects of viscosity and resolution

In the previous section, we discussed the properties of the energy transfer and
inertial dissipation around an exact singularity. How is this picture changed for a
867 P1-33
B. Dubrulle
(a) (b) d¶I/´

1.0 0.10 20
0.08
0.5 15
0.06
u(x)
0 ¶/L 10
0.04
-0.5 5
0.02
0
-1.0 -0.5 0 0.5 1.0 -1.0 -0.5 0 0.5 1.0
x x
F IGURE 18. (a) Velocity fields uν (x, t) at t = 1 for solution of the Burgers equation for
different viscosity in a domain of size 2L = 2, with shock amplitude 1u = 2. Blue symbols:
ν = 1/100; green symbols ν = 1/10. The red line is the Khokhlov sawtooth solution,
corresponding formally to ν = 0. (b) Space/scale diagram of the non-dimensional local
energy transfer DÌ / for the Khokhlov sawtooth solution. The intensity of DÌ / is coded
in colour, following the colour bar.
(a) 10 4 (b) 1.2

-1
¶ 1.0
10 2
0.8
d I¶(0)/´
¯d I¶˘/´
10 0 0.6
0.4
¶2
10-2
0.2
10-4
10-3 10-2 10-1 0 0.02 0.04 0.06 0.08 0.10
¶/L ¶/L
F IGURE 19. Properties of the local energy transfers for the Khokhlov (red) and viscous
solutions (blue: ν = 1/100; green: ν = 1/10) of the Burgers equation shown in figure 18(a).
(a) DÌ (0)/ as a function of scale `. Note that DÌ (0)/ is the maximum of DÌ (x)/ over
the interval [−L, L]. (b) Average of DÌ (x)/ over the interval [−L, L] as a function of
scale `.
quasi-singularity, i.e. a singularity that is smoothed (regularized) by the introduction

of viscosity? To answer this question, we can again take the shock as an illustration.
We consider the energy transfer for the finite viscosity solution uν (x, t) for two
values of ν, corresponding to a very smooth solution (large viscosity) and a
nearly singular solution (small viscosity), for which the length of variation of
867 P1-34
Beyond K41
(a) d¶I/´ (b)

0.10 2.0
100 ¶-1
0.08 1.5
d I¶(0)/´, d ˜¶(0)/´
0.06
¶/L 1.0
10-2
0.04
0.5 ¶-2
¶2
0.02
0
-1.0 -0.5 0 0.5 1.0 100
x ¶/˙s
F IGURE 20. (a) Space/scale diagram of the non-dimensional local energy transfer DÌ /
for the viscous solution of the Burgers equation with ν = 1/10. The intensity of DÌ / is
coded in colour, following the colour bar. (b) Universal local energy budget as a function
of the scale for the viscous solution of the Burgers equation with ν = 1/10 (green symbols)
and ν = 1/100 (blue symbols). Square: DÌ (0)/; triangle: D`ν (0)/.
the velocity near the jump has a size comparable to the resolution, see figure 18(a).
The space/scale diagram for the local energy transfer DÌ for these two solutions
depends on the viscosity. For the nearly singular solution, it is very similar to the
space/scale diagram of the singular solution (not shown). For larger viscosity, one
starts observing differences, as illustrated in figure 20(a): the intensity of the signal
is smaller, and decreases towards zero at sufficiently small scale. A quantification
of this effect can be obtained by looking at the values of the inertial dissipation
at the shock, represented by DÌ (0) and the mean value of the inertial dissipation,
represented by hDÌ i. They are reported in figure 19 for the two cases. We see that
in both cases, DÌ (0) tends to 0 as ` → 0, meaning that the inertial dissipation is
zero. This is natural, since both solutions are regular (see § 6.2.3). Also, looking
at hDÌ i, we see that it converges to zero like `2 for the more regular solution.
The less regular solution does not satisfy this scaling, because one would need a
resolution smaller than the size of the velocity variation to observe this. However,
DÌ (0) still follows the 1/` scaling for large enough scales, like the singular solution.
There is therefore a ‘footprint’ of the singularity at finite but small enough scales, as
discussed in Saw et al. (2016), which opens the pathway to ‘scan’ quasi-singularities
of the Navier–Stokes equations by tracking the energy transfers. Such observations
motivate singularity or quasi-singularities detection in turbulent flows using DÌ , as
discussed in Kuzzay et al. (2017), Saw et al. (2016).
The variations of DÌ with scale seem to depend strongly on viscosity. However,
there is a way to collapse all viscous solutions onto universal curves by introducing
the typical length ηs = ν/1u, which is the only characteristic scale one can build
867 P1-35
B. Dubrulle
(a) V⊥ (b)
1.0
0.6
0.4
0.5 12
0.2 10
8
¶/Îx
0 0 6
z/R 4
-0.2 2
1
-0.5 1
-0.4
0
0
-0.6 z/R
-1.0 -1 -1 x/R
-0.5 0 0.5
x/R
-4 -2 0 2 4
d¶I/´
F IGURE 21. (a) Example of an instantaneous velocity field, measured in a plane

containing the rotation axis, in the von Kármán experiment with parameters similar to
experiment ANTIG, at resolution 1x = 1.2 mm. The arrows code the in-plane velocity,
the colour codes the out-of-plane velocity, V⊥ . The velocities and axis are non-dimensional.
The units of length and time are taken as R, the cylinder radius, and 1/2πF, the rotation
period of the impellers. (b) Example of an instantaneous measurement of the local energy
transfer DÌ in experiment ANTIG, as a function of space and scale, in a plane containing
the rotation axis. The local energy transfer is expressed in units of and is coded with
colour. The scale is expressed as a unit of the spatial resolution of the measurements.
using the shock amplitude and the viscosity. Expressing all scales in units of ηs , we
then get figure 20(b) for DÌ (0). One sees that the curves corresponding to different
viscosities now collapse, scaling like `−1 for ` > ηs and `2 for ` < ηs . The same
universality also holds for the viscous term D`ν (0) that now scales like `−2 for ` > ηs
and tends towards for ` < ηs . The location where ` = ηs is precisely the location
where the ‘inertial’ term DÌ (0) balances the viscous term D`ν (0). These observations
are one of the basic ingredients of the generalization of the Kolmogorov picture
of turbulence, described in § 7. The other ingredients result from the link between
WKHM and intermittency in three-dimensional turbulent flows.
6.3. WKHM and intermittency

6.3.1. Spatial distribution of energy transfers and dissipation
Let us go back to real turbulence and try to compute DÌ and D`ν on a velocity
field measured in the von Kármán flow. An example is shown in figure 21, for
scales of the order of ` ∼ 100/η. We see that DÌ is inhomogeneous in space. In
scale, the maxima of DÌ are distributed along thinning lines that point toward
the location of maxima of DÌ at the resolution scale 1x = 120η. This picture is
reminiscent of what we observe for the Burgers velocity field, where we know that
867 P1-36
Beyond K41
(a) d¶I/´ (b) d¶˜/´

0.10 2.5
0.05 0.05
0.05 2.0
1.5
y 0 0 0
1.0
-0.05
-0.05 -0.05 0.5
-0.10 0
-0.05 0 0.05 -0.05 0 0.05
(c) d¶I/´ (d) d¶˜/´

0.3 4
0.05 0.2 0.05

3
0.1
y 0 0 0 2
-0.1
1
-0.05 -0.2 -0.05
-0.3
0
-0.05 0 0.05 -0.05 0 0.05
x x
F IGURE 22. Examples of topology around an extreme value of the local energy transfer
DÌ at the dissipative scale ` ≈ η in experiment CONC-3D. Around the event, coded in
colour, the velocity field, shown by white arrows, displays a ‘shock-like’ (a) or ‘spiral-
like’ (c) structure. In (b) and (d), the local viscous energy dissipation D`ν for the two
types of events is provided, coded by colour. The white contour delimitates the area A1.
The figures are drawn using unpublished data collected by the EXPLOIT collaboration
(F. Daviaud, B. Dubrulle, P. Debue, V. Valori, J.-P. Laval, J.-M. Foucaut, Ch. Cuvier, Y.
Ostovan), with permission. Adapted from Debue (2019).
the point of convergence of the thickening lines of maxima is the shock i.e. the
velocity field singularity regularized by viscosity at the Kolmogorov scale. To check
that this picture is also valid in the von Kármán case and to study properties of
corresponding quasi-singularities or singularities, we need to zoom into the flow, to
decrease ` down to the Kolmogorov scale.
Two examples are shown in figure 22, where the spatial distributions of D`ν and
DÌ are plotted around a location of convergence of maxima lines in scale space
for the experiment ANTIC-3D, where the smallest accessible length is of the order
of the Kolmogorov scale. D`ν is always positive, while DÌ can take both positive
and negative values, tracing both downscale (positive) and upscale (negative) energy
867 P1-37
B. Dubrulle
(a) log10(¶/˙) (b)

2 3.0
0
0 2.5
-2 2.0
-5
log10(P)
-4 1.5
-6 1.0
-10
0.5
-8
0
-10 -15
-200 0 200 -5 0 5
(d I¶ - ¯d I¶˘)/ß (ln(d ˜¶) - ¯ln(d¶˜)˘)/ß
F IGURE 23. Centred and reduced PDFs of local energy transfers DÌ (a) and log of
local viscous dissipation D`ν (b) at different scales, coded by colour, for experiments
ANTIG and ANTIC-1 to ANTIC-4 (see figures 7(a) and 27(b) for corresponding mean
and standard deviation and table 1 for symbols).
transfers. At this scale, downscale energy transfers clearly dominate, however, as

one goes towards larger and larger scales, this difference weakens and upscale
energy transfers become more numerous (Debue et al. 2018). One sees that the
largest values of DÌ are located in a coherent structure, that is connected with
a shock structure in the velocity field. Outside such structures, the value of DÌ
is much smaller. The viscous dissipation D`ν is also highest within the coherent
structure, but does not follow exactly the pattern of the local energy transfers DÌ .
It conserves significant values in the region to the right of the shock, while it is
negligible in the region left of the shock. This kind of structure might therefore
be associated with the roll up of a vorticity sheet viewed from above. They seem
to be the most frequent structures encountered in the von Kármán flow (Saw et al.
2016). The shocks are not the only coherent structures that are associated with large
values of DÌ . The latter can also be associated with a ‘spiral’-like structure in the
velocity (a vortex), as illustrated in figure 22(b). In this case, the energy transfer
DÌ can be both positive and negative within the core of the vortex and negligible
elsewhere, while the viscous dissipation D`ν is strong within the core, and extends
outside, forming spiral arms. Such a structure could therefore be associated with
the reconnection of two vortex tubes or the roll up of a vortex sheet viewed from
the side. Other types of coherent structures, observed less frequently, include ‘cusps’
and ‘jets’ (Saw et al. 2016).
6.3.2. Statistical properties of local energy transfers and dissipation

The spatial distribution of local energy transfers and dissipation is very
inhomogeneous, at all scales. This is reflected in their statistics, shown in figure 23,
867 P1-38
Beyond K41
(a) 106 (b) 103

/(Í1DR)p/3˙†(p/3)
/(Í1˜)p/3˙©(p/3)
104 102
102 101
Íp/3
Íp/3
DR
˜
100
100
100 101 102 103 100 101 102 103
¶/˙ ¶/˙
F IGURE 24. Scale variation of the non-dimensional structure function of order p = 1/3
to p = 2 of local energy transfers and dissipation for experiments ANTIG and ANTIC-
1 to ANTIC-4. The structure functions have been shifted by arbitrary factors for clarity
and are coded by colour: p = 1/3: blue symbols; p = 2/3: orange symbols; p = 1: yellow
symbols; p = 4/3: magenta symbols; p = 5/3: green symbols; p = 2: light blue symbols.
(a) Non-dimensional structure functions for DÌ . (b) Non-dimensional structure functions
for D`ν . The dashed lines are power laws with exponents shown in figure 13(b). Adapted
from Debue et al. (2018).
to be compared to the statistics of the averaged energy transfers, studied in

§ 3.4 (figure 6). The distributions are similar in shape in both cases, but the
tails of the distributions are fatter at small scale in the case of local energy
transfers and dissipation. As the length is increased, the distributions tend to look
increasingly like the distributions of the average, which is natural since DLI = ΠLI
and DLν = ΠLν . The self-similarity is broken, so that exponents of the structure
ν
functions ΣpDR = h|DÌ |p i ∼ `ξ ( p) and Σpν = h|D`ν |p i ∼ `ξ ( p) , shown in figure 24, do
I
not follow a simple linear law, as can be seen on figure 13(b).
6.3.3. Scaling of energy transfers and dissipation

When viewed in the scale space, the behaviour of energy transfers and dissipation
also reveals an interesting difference between regions outside or within ‘coherent’
structures. Such structure are identified as connected regions where energy transfers
have very strong values, and are delimited by a white line in figure 22. Now, we
can perform 3 kinds of conditional average: (i) average of DÌ and D`ν inside the
coherent structure (A1); (ii) average over the regions corresponding to the areas
where DÌ takes values larger than seventy per cent of its maximum over the whole
field (A2, strong energy transfers); (iii) average over the regions corresponding to
the areas where DÌ takes values of less than ten per cent of its maximum over the
whole field (A3, weak energy transfers). In choosing the threshold for the two cases,
we have ensured that they include the same number of points (approximately 70),
867 P1-39
B. Dubrulle
(a) (b) 101
100 100
¯d I¶˘/´, ¯d ˜¶˘/´
¶-4/3
10-1
¶-4/3
10-1
¶2 ¶2
10-2
10-2 10-3
100 100
¶/˙ ¶/˙
F IGURE 25. Local energy budget as a function of the scale for the two velocity fields
of figure 22, average over different regions, coded by symbols: circle: A1/structure
region; stars: A2/weak transfer region; triangles: A3/strong transfer region. (a) Shock case;
(b) spiral case: DÌ (0)/ averaged over region A1 (magenta circle), A2 (red stars), A3
(light blue triangles): D`ν (0)/ averaged over region A1 (green circle), A2 (yellow stars)
and A3 (blue triangles). The figures are drawn using unpublished data collected by the
EXPLOIT collaboration (F. Daviaud, B. Dubrulle, P. Debue, V. Valori, J.-P. Laval, J.-M.
Foucaut, Ch. Cuvier, Y. Ostovan), with permission.
for a detailed comparison. The scaling of these conditional averages are shown
in figure 25(a) for the shock case, and figure 25(b) for the spiral case. In all
situations we observe similar trends: the term hDÌ i decreases with decreasing scale,
with a clear change of slope at the point where it crosses the value of hD`ν i in
the case of weak energy transfers. The term hD`ν i increases for decreasing scales,
and then saturates towards a plateau, which defines the local energy dissipation in
the corresponding region. We call this value for the weak energy transfers case
(A3), and normalize all curves using this value, and its corresponding Kolmogorov
scale η = (ν 3 /)1/4 . Note that is higher than g , the global energy dissipation,
corresponding to the total dissipation within the flow, and measured by torquemeters:
about twenty per cent higher in the case shown in figure 25(a), and 90 per cent
higher in the case shown in figure 25(b). Then, we find that dissipation is 5 times
larger than within the spiral (A1 region case (b)), while it 2 times higher than
within the shock (A1 region case (a)). Overall, the three curves look similar with
the curves of the K41 theory (figure 4a), with a scaling close to `2 below η for
hDÌ i and `−4/3 above η for hD`ν i. A closer inspection shows that cases A1 and A2
behave identically with respect to A3: the corresponding hD`ν i are systematically
steeper than for case A3 when ` > η and hDÌ i are systematically shallower than for
case A3. Finally, the crossing between hD`ν i and hDÌ i occurs at a scale smaller in
cases A1 and A2 than in the case A3. Overall, the case A3 seems to be closer to
the K41 picture, meaning that the turbulence is more ‘K41-like’ within the regions
867 P1-40
Beyond K41
of moderate local energy transfers and more ‘non-Kolmogorov’ within regions
where local energy transfers are strong (A2 and A1 regions). Can we strengthen
this affirmation?
6.3.4. Intermittency and energy transfers

Let us see how the strong energy transfers connect with intermittency, by
conditioning the structure functions on the regions of high or low local energy
transfers at the resolution scale. We implement this for the each data set by defining
a special set of points A corresponding to the locations of low of |D1x I
|. To define
it, we divide the PDF of |D1x | at the smallest scale of a given experimental data
I
set into 10 sets: the set of points forming the first decile (spatial regions with |D1xI
|
being in the 10 per cent lowest values of the inertial dissipation) are assigned to
the set A.
We compute the PDF of the wavelet velocity increments conditioned on the set A.
It is shown in figure 10(b). With respect to the non-conditioned case, (figure 10a),
we see that the tails are slightly truncated, especially at smaller scales. This means
that the large wavelet velocity increments are correlated with large events of D1x I
,
especially at small scale. This effect is more obvious when reducing and centring
the distribution (inset figure 10b1), which looks however not much more self-similar
than the unconditioned one. We now turn to the scaling behaviour of the mean and
of the std of the conditioned distributions, plotted in figure 11(a,b).They appear to
follow the `1/3 as well as their unconditioned counter-part. A more stringent test of
the quality of the self-similarity is given by looking at the ratio of the std to the
mean, provided in figure 8(b). If the distribution is self-similar, this ratio should
be constant. We see that this is not true for the unconditioned case, while it is
approximately true for the conditioned case.
A more refined test of the self-similarity is provided by the scaling wavelet
velocity structure functions. We compute them as
SnC (`) = h|δW(u)(x, `)|p iA , (6.28)
where the average is taken on the set A. We show such structure functions in
figure 12(b) for orders p ∈ [1, 6]. From them, we compute the scaling exponent
ζC (p). They are shown in figure 13(a). We see that ζC (p) is almost straight, lying
very close to the Kolmogorov K41. The intermittency is reduced, meaning that
the regions corresponding to the set A are more self-similar. This shows that
the intermittency is connected to the high events of the local energy transfer at
the resolution scale. This suggests that the natural analogue of the quantity ` in
K62, is DÌ (Drivas & Eyink 2017; Debue et al. 2018) and that all intermittency
corrections can be understood using the statistical scaling properties of the local
energy transfers and dissipation defined through WKHM. The proper generalization
involves the multi-fractal (MFR) formalism, first proposed by Frisch & Parisi
(1985).
867 P1-41
B. Dubrulle
7. Beyond Kolmogorov using multi-fractals
There are many excellent reviews on the multi-fractal theory, see e.g. Chevillard
(2004), Boffetta, Mazzino & Vulpiani (2008), Benzi & Biferale (2009), Frisch
(2016). Here, I summarize only the notions that are linked with the WKHM
picture.
7.1. The multi-fractal model

7.1.1. Self-similar blow-up solutions
There is now evidence for the existence of self-similar solutions (SSs) for
several nonlinear systems, with the same symmetries and conserved quantities
as Navier–Stokes. These solutions are invariant under the scaling symmetry
(t, x, u) → (λ1−h t, λx, λh u) and satisfy
)
u(x, t) = (t∗ − t)h/(1−h) F(x/(t∗ − t)1/(1−h) ), backward solution,
(7.1)
u(x, t) = (t − t∗ )h/(1−h) F(x/(t − t∗ )1/(1−h) ), forward solution,
where t∗ is the blow-up time and h < 1 has the meaning of a Hölder exponent. The
backward solutions may describe a singularity formation, while the forward solutions
may be associated with non-uniqueness of associated weak solutions. An example
of backward SS is the Khokhlov sawtooth solution with L → 0, a solution with h =
0 and t∗ = 0 (see Eggers & Fontelos (2009) for more examples). In inviscid shell
models of turbulence, the backward SSs were studied by Dombre & Gilson (1998),
Mailybaev (2012, 2013). The velocity U(kn , t) in the shell of wavenumber kn then
takes the form (7.1) with h = 0.281 and x being replaced by 1/kn . In the presence
of viscosity, these solutions disappear, but the system ‘keeps the memory’ of their
existence via extreme events of velocity or velocity derivatives, correlated over space
and time, sometimes named ‘instantons’ (Mailybaev 2013). Such a memory effect
is also clearly seen for the Burgers solution in the scale/space diagram figure 20(a),
where both the velocity and the local energy transfers are highest at the location of
the inviscid shock. Moreover, the ideal inviscid scaling of the Khokhlov solution is
still valid for scale above the dissipative scale (figure 20b).
The existence of SSs for Euler or Navier–Stokes equations is still controversial.
In the Euler case, Chae proved that there is no backward SS (Chae 2007), but
self-similar blow-up behaviours with h = −1 were found on logarithmically spaced
lattice (Campolina & Mailybaev 2018), or during reconnection of tent vortices
(Kimura & Moffatt 2018). In the Navier–Stokes case (see Bradshaw & Tsai (2018)
for a recent review), the SS necessarily has an h = −1 exponent (in agreement with
the h = −1 rescaling symmetry, see § 4.1) and blows up in finite time. The backward
case was first investigated by Leray (1934) and was excluded by Necas, Ruziicka
& Sverak (1996) for a large class if initial conditions. It has however recently
been claimed to be observed in high speed wind tunnel experiments (Leberre &
867 P1-42
Beyond K41
Pomeau 2018). In the forward case, the existence of SS solutions is known under
certain regularity conditions. Nevertheless, both backward and forward SS have
been observed for energy spectra in the Leith model of turbulence (Nazarenko
& Grebenev 2017). The backward solution corresponds to an initial explosive
propagation of the spectral front from the smallest to the largest wavenumbers
reaching arbitrarily large wavenumbers in a finite time, and is described by a
self-similar solution of the second kind. The forward solution manifests itself as a
reflection wave in the wavenumber space propagating from the largest toward the
smallest wavenumbers, and is described by a self-similar solution of a new (third)
kind.
7.1.2. The multi-fractal phenomenology

Even though the SS solution may not be relevant for the Euler or Navier–Stokes
equations, they provide an interesting example of velocity fields with Hölder
exponent h since for fixed time, their wavelet transform scale like `h . Note
finally that in the case of Euler equations, some weak solutions with Hölder
exponent h < 1/3 have recently been constructed by Buckmaster et al. (2018),
which provides further support for the idea that, in the inviscid limit, there are
solutions characterized by velocity increments or wavelet coefficient scaling like `h ,
for arbitrary small `. The multi-fractal model of Frisch & Parisi (1985) takes one
step further and assumes that, in presence of viscosity, there exists a whole family
of locally self-similar solutions, labelled by a scalar field h(x, t, r), such that
ln |δr u(x, t, r)|
h(x, t, r) = , ` = |r| (7.2)
ln(`/L)
for a range of scale in a suitable ‘inertial range’ ηh ` L, where L is a
characteristic integral length of scale and ηh a cutoff length scale. Such a definition
of h is mathematically sound only for 0 6 h 6 1. Using wavelet coefficients δW(u)
instead of velocity increments, one can extend the definition of h to values h < 0
(Muzy et al. 1991; Eyink 2007–2008). The local energy balance for such solutions
helps us to understand the meaning of the cutoff scale ηh . Indeed, plugging (7.2)
into the definitions (6.14) and (6.15), we see that DI` ∼ `3h−1 and Dν` ∼ ν`2h−2 .
These two terms balance at a scale ηh ∼ ν 1/(1+h) . This scale thus appears as a
fluctuating cutoff which depends on the scaling exponent and therefore on x. This
is the generalization of the Kolmogorov scale η1/3 ∼ ν 3/4 , and was first proposed
in Paladin & Vulpiani (1987). Below ηh , the viscous dissipation takes over and
regularizes the flow, exactly like it did in the K41 solution or in the Burgers shock
solution. The velocity increments then scale like |δu(x, `)| ∼ ` so that DI` decreases
to 0 like `2 , while Dν` tends to a constant, equal to the local energy dissipation
. This picture is thus a generalization of K41, and is shown schematically in
figure 4(b). We note that the local energy transfer of the viscous Burgers solution
867 P1-43
B. Dubrulle
(figure 20b), exactly follows such scheme, with h = 0 being the Hölder exponent of
the solution at zero viscosity, resulting in νs ∝ ν.
7.1.3. The multi-fractal and Onsager’s dissipative solutions

As long as h > −1, we have limν→0 ηh = 0. In that limit, the wavelet coefficients
δW(u, `) are self-similar with exponent h at all scales, and the corresponding
solution tends to a solution of Euler equation with Hölder of exponent h. In this
respect, we can interpret the corresponding multi-fractal solutions as (regularized)
footprints of solutions of the Euler equation with a Hölder of exponent h. As
we have seen in § 6.2.4, such solutions are able to dissipate energy for the Euler
equation if −1 < h 6 1/3. For this reason, we call multi-fractal solutions with
−1 < h 6 1/3 dissipative quasi-singularities, while the solutions with 1/3 < h 6 1
will be called non-dissipative quasi-singularities.
The multi-fractal solution for h = −1, if it exists, may be considered as an
equivalent of the Leray SS of the Navier–Stokes equations. It is quite special, since
η−1 = limh→−1 Re−1/(1+h) = 0 at any fixed Re > 1. For this exponent, there is no
possibility of regularization by viscosity (1/r is a zero mode of the Laplacian),
so that h = −1 would correspond to a Navier–Stokes singularity which dissipates
energy by non-viscous means.
The natural question that occurs now is: how do these different solutions
contribute to the total energy dissipation, and can they explain the dissipation
anomaly? To answer this question, we must first be able to count how often each
of them occurs within the flow, i.e. what is their probability of occurrence. Does it
mean that we have to be able to measure, with good precision, the velocity field
at each point, compute its wavelet transform, fit locally an exponent and form a
histogram? Physically speaking, the procedure looks insane, because of noise issues.
Mathematically speaking, it would be a complete nonsense, since the exponents
h(x) may be only defined on sets of zero measure (see e.g. Caffarelli, Kohn &
Nirenberg 1982). Fortunately, there exists a procedure to define the ‘probability of
h’, based upon the large deviation theory.
7.1.4. Multi-fractal and large deviations

Since the velocity field is turbulent, h ≡ ln δW(u, `)/ ln(`/L) varies stochastically
as a function of space and time. Also, if the turbulence is statistically homogeneous
and isotropic, h only depends on `, the scale magnitude. Therefore, formally, h
can be seen as a stochastic process labelled continuously by ln(`/L). By Kramer’s
theorem (Touchette 2009), one sees that in the limit ` → 0, ln(L/`) → ∞ and we
have
`
C(h)
ln(`/L)C(h)
,

Prob ln(δW(`)) = h ln(`/L) ∼ e = (7.3)
L
867 P1-44
Beyond K41
where C(h) is the rate function of h, also called the multi-fractal spectrum. Formally,
C(h) can be interpreted as the co-dimension of the set where the local Hölder
exponent at scale ` is equal to h. This observation allows us to single out the
exponents where C(h) > d, the space dimension. Indeed, such exponents correspond
to solutions that are so rare that the probability of observing them in a finite volume
during a finite time interval is negligible.
Using Gärtner–Elis theorem (Touchette 2009), one can connect C and the wavelet
structure functions as
hδW p i = heph ln(`/L) i ≈`→0 eln(`/L)λ( p) , (7.4)
where λ(p) is the Legendre transform of the rate function, i.e. λ(p) = minh (ph +
C(h)). Since eln(`/L)λ( p) ∼ `λ( p) , we see that λ(p) = ζ (p), where ζ (p) is the scaling
exponent of the wavelet structure function. We can then compute the probability
of finding a given exponent h by performing a Legendre transform on these ζ (p).
Because C(h) is a Legendre transform, it is necessarily convex. The set of points
where it satisfies C(h) 6 d (representing the set of admissible, or observable h) is
therefore necessarily an interval, bounded by −1 6 hmin and hmax 6 1.
7.2. Illustration: the Burgers solution

The multi-fractal spectrum can be computed analytically in some simple cases, such
as the Khokhlov solution of the inviscid Burgers equation (6.27). The structure
function of order p is given by (Eyink 2007–2008)
` ` ` 2L + ` p
p
p
h|δu| i = 1 − + . (7.5)
2L t 2L t
When p < 1, the first term dominates and is equal to (1u)p (`/2L)p when ` L so
that ζ (p) = p. When p > 1, the second term dominates and is equal to (1u)p (`/2L)
when ` L, so that ζ (p) = 1. Taking the inverse Legendre transform, for p < 1, we
get C(1) = 0: the set points with Hölder exponent h = 1 (the regular points) indeed
covers almost all the interval [−L, L]. Taking the inverse Legendre transform for
p > 1, we get C(0) = 1: the location where h = 0 (the shock) corresponds to the
point x = 0, which is a set of co-dimension 1. In this example hmin = 0 and hmax = 1.
When we consider the viscous solution instead, the scaling of the structure
function given by (7.5) only holds for ` > ηs . Below ηs , δu ∼ ` because of regularity,
and the structure function of order p scales like `p (figure 26a). The multi-fractal
spectrum C(h) can nevertheless be computed using the scaling in the ‘inertial range’
` > ηs (figure 26b) and provides the value of the ideal ‘inviscid’ case (figure 26b1).
This case illustrates the methodology that is applied in real turbulence data.
867 P1-45
B. Dubrulle
(a) (b) 1.5
1.0
¯|∂u|p˘/(Îu)p
100
†B(p)
(b1) 1.5
0.5
1.0
CB(h)
¶†B(p) 0.5
0
0
p -0.5
10-5 ¶ -0.5 0 0.5 1.0 1.5
h
-2 0 -0.5
10 10 0 1 2 3 4
¶/˙s p
F IGURE 26. Multi-fractal analysis of the Burgers solutions. (a) Scale variation of the non-
dimensional wavelet structure function computed using velocity fields uν (x, t) of figure 18
with ν = 1/100 (open symbols) and ν = 1/10 (filled symbols). The structure functions
have been shifted by arbitrary factors for clarity and are coded by colour: p = 1/3: yellow
squares; p = 2/3: red stars; p = 1: magenta diamonds; p = 4/3: black circles; p = 5/3:
blue stars; p = 2: blue triangles; p = 7/3: green triangles. Note that because the structure
functions scale like `p in the viscous range ` < ηs , there is no scaling difference below
and above ηs for p 6 1. (b) Scaling exponents τB as a function of order p. Inset (b1):
multi-fractal spectrum CB (h) obtain by taking inverse Legendre transform of τB (p).
7.3. Application of the multi-fractal model to turbulence

7.3.1. Interpretation of extreme events of local energy transfer
We have now tools at hand to interpret the behaviour of local energy transfer and
dissipation, measured in the von Kármán flow (§ 6.3). Typically, we observed that
local energy transfers are very intermittent in space and time. Even at scales close to
or below the Kolmogorov scale, they exhibit locally intense events, than can exceed
the global energy dissipation by several orders of magnitude (see figure 22 and
Saw et al. (2016)). In the multi-fractal picture of turbulence (figure 4b), such events
correspond to dissipative quasi-singularities (h 6 1/3), that have not yet reached
their dissipative cutoff ηh . The local energy budget of figure 25 can then be viewed
as a (noisy!) realization of figure 4), with h ∼ 0.2, obtained by fitting the scaling
of Dν` above the viscous cutoff. In contrast, the location where the local energy
transfers are small above the Kolmogorov scale, correspond to non-dissipative quasi-
singularities (h > 1/3). The dissipative quasi-singularities with −1 < h 6 1/3 all
eventually convert their local energy into viscous dissipation at ηh , the corresponding
energy being transferred at a nearly constant rate towards smaller and smaller scales
by the viscous term Dν` , until it is ultimately converted into heat. In such picture, the
value of the local energy dissipation is therefore provided by
= Dνηh = DIηh . (7.6)

867 P1-46
Beyond K41
Using the scaling DI` ∼ `3h−1 , and ηh ∼ Re−1/(1+h) , we thus get ∼ Re(1−3h)/(1+h) ,
which increases as Re increases or h decreases. This picture therefore predicts that
the fluctuations of the energy dissipation become larger and larger and result from
events occurring at smaller and smaller scale as Re is increased. It also means that if
one truncates a numerical simulation at a scale 1x larger than ηhmin , the cutoff scale
of the smallest Hölder exponent that can be observed in the flow, then one misses all
the events of energy dissipation produced by quasi-singularities with hmin 6 h 6 −1 −
ln Re/ ln(1x/L). This explains why, increasing a numerical simulation from 1x =
η/2 to η/10, the tails of the energy dissipation increase, as noticed by Yeung, Zhai
& Sreenivasan (2015). The amount of energy dissipation lost by lack of resolution
actually depends on hmin and on C(h) that can either be measured (see § 4.5), or
analytically computed on simple models, such as the Burgers solution (§ 7.2).
7.3.2. Scaling range

The previous discussion shows that the notion of non-trivial Hölder exponent h <
1 is only valid provided ` > ηh . The notion of a multi-fractal is thus rigorously
only meaningful in the Re → ∞ limit, where the limit ` → 0 can be taken. It is
indeed in this limit only that all the properties deriving from large deviations, and
the Legendre property (such as convexity, etc.) are valid. In the following, we will
assume that Re is sufficiently large so that there exists an ‘inertial range’ of scale
η1/3 ` L, in which the multi-fractal formalism applies. Since ηh is a decreasing
function of h, this ensures that the condition ηh ` holds for all h 6 1/3 i.e. for all
dissipative solutions.
7.3.3. Scaling exponents and signs of DÌ and D`ν

Once the multi-fractal spectrum C(h) is given, one can use the small-scale
limits of § 6.2.3 to compute the scaling exponents of the wavelet velocity structure
functions via ζ (p) = minh (ph + C(h)), but also the scaling of other quantities
appearing in WKHM. For example, since D`ν ∼ `2h−2 and DÌ ∼ `3h−1 , we have
ν
h|D`ν |p i ∼ `ξ ( p) , ξ ν (p) = minh (p(2h − 2) + C(h)),
)
(7.7)
h|DÌ |p i ∼ `ξ ( p) , ξ I (p) = minh (p(3h − 1) + C(h)).
I
Using the regularity of the velocity at scale ` = ηh ∼ Re−1/(1+h) , we can also get
useful scalings with Reynolds number. Indeed, at ` = ηh , we have δr u(x, t) = rj ∂j ui
so that we derive the following useful trace formula:
Dηνh ∼ νTr(SS + ),
)
(7.8)
DηIh ∼ ηh2 Tr(S + S 2 ),
where S ij = ∂j ui . From this, we get a connection between the sign of the inertial
or viscous energy transfers, and the topology of the coherent structures at the
867 P1-47
B. Dubrulle
dissipative scale. Indeed, we have Tr(SS + ) = i=1,3 |λi | , and Tr(S S ) =
2 + 2
P
i=1,3 λi |λP
i | , where λi are the eigenvalues of the tensor S . Using the incompressibility
2
P
i λi = 0, we see that (i) Tr(SS ) is positive; (ii) if the eigenvalues of

+
condition
S are real with λ1 > λ2 > λ3 , then Tr(S S ) = 3λ1 λ2 λ3 . This means that Dηνh , is
+ 2
positive and that DηIh is zero for quasi two-dimensional flows (λ2 = 0) or is positive
for filaments-like structures (λ2 > 0), and negative for sheet-like structures (λ2 < 0).
7.3.4. Constraints on the multi-fractal spectrum

Turbulence properties impose some constraints on the multi-fractal spectrum C(h).
They are:
Normalization. The normalization of the PDF imposes h|δW|0 i = 1, so that
ζ (0) = min(C(h)) = 0. (7.9)
h
Let us call h∗ the value that realizes the minimum C(h∗ ) = 0. It corresponds to the
exponent reached with the highest probability, i.e. the most probable exponent. In
K41, we would have h∗ = 1/3. However, intermittency effects produce a shift in
this most probable exponent, as we illustrate in § 7.3.5.
Kolmogorov’s 4/3 law. Equation (3.2) implies ζ (3) = 1. This implies C(h) > 1 −
3h, the equality being reached by the h1 obeying
dC/dh|h1 = −3, for h1 s.t. C(h1 ) = 1 − 3h1 . (7.10)
Note however that this constraint strictly applies only to quantities built from
ingredients that are involved in the 4/3rd law, i.e. third power of velocity increments,
with no absolute value.
Anomalous dissipation. The condition hi = cte for Re → ∞ imposes χ (1) = 0.
Since 1 + h > 0, this is achieved if C(h) > 1 − 3h, equality being achieved for h = h1
defined via (7.10): the 4/3th law and the anomalous dissipation provide the same
constraint on the multi-fractal spectrum. Since the 4/3th law is directly derived from
the Navier–Stokes equation, one therefore concludes that the anomalous dissipation
is a consequence of the 4/3th law and built into the Navier–Stokes equations.
7.3.5. Illustration: the log-normal model

The simplest non-trivial convex function with a minimum equal to 0 is a parabola
C(h) = (h − a)2 /2b, which satisfies the normalization condition (7.9). Performing the
Legendre transform, we see that the corresponding structure function exponents are
given by
ζ (p) = ap − bp2 /2, (7.11)
so that it corresponds to a log-normal law for the underlying process.
√ In this case,
√ probable exponent in dimension 3 is h = a, while hmin = a − 6b and hmax =
the most
a + 6b.
867 P1-48
Beyond K41
Such a multi-fractal spectrum depends on two parameters a and b. To be a valid
approximation for turbulence, it must satisfy the constraints of § 7.3.4. The 4/5th law
condition (7.10) imposes a = 1/3 + 3b/2 so that the multi-fractal spectrum is given
by
(h − 1/3 − 3b/2)2
C(h) = . (7.12)
2b
One sees that the most probable exponent is shifted from 1/3 by a term proportional
to b, which therefore encodes all intermittency corrections. One more constraint
only is needed in this approximation, to compute the intermittency parameter. For
example, if one imposes hmin = 0, so that the strongest dissipative solution has the
same Hölder regularity as a shock, one gets b ≈ 0.023. On the contrary, if one
imposes that hmax = 1, so that the multi-fractal spectrum extends all the way to
the regular solutions, one gets b ≈ 0.056. In the next section, we use our data to
compute C(h) from scaling exponents and get estimates of b.
7.3.6. Measurement of C(h) and consequences

For a given set of ζ (p), we can perform a Legendre transform and get the multi-
fractal spectrum C(h). Doing that on the exponents of figure 13(a), we obtain the
result provided in figure 27(a). One sees that it parabolic, and is well fitted by a
parabola of the shape (7.12), with b = 0.045 and a = 0.35. We note that this does
not satisfy a = 1/3 + 3b/2, because indeed, in our data, ζ (3) ≈ 0.8 (remember that
we have considered absolute values of wavelet increments that are not involved in
the 4/3rd law). Extrapolated up to C = 3, we get hmin ≈ −0.2.
We can also substitute this value into (7.11) to check that it provides a good fit
of the exponent ζ (p), shown in figure 13(a). We can also use them to compute the
scaling exponent of the generalized skewness Sp given by Sp = h|DÌ |p i/h|D`ν |3p/2 i.
This exponent is ξ I (p) − ξ ν (3p/2) = 2p, i.e. the same value found in § 4.2 for
(ΠÌ )q /(Π`ν )3q/2 .
More stringent test can be done by considering other scaling exponents introduced
previously. Indeed, in the log-normal approximation, we have
)
ξ ν (p) = − 43 p + bp(3 − 2p),
(7.13)
ξ I (p) = 92 bp(1 − p).
These values are used to compute ξ ν (p/3) − p/3 and ξ I (p/3) − p/3 and have been
reported in figure 13(b). We see that they do not provide a good fit of the data,
which seem better approximated by formulae (7.13), with b = 0.065. How can we
understand this?
7.4. WKHM multi-fractal refined similarities

From figure 13(b), we see already a striking feature: the intermittency exponent of
velocity increment τ (p/3) = ζ (p) − p/3 does not coincide with the scaling exponent
867 P1-49
B. Dubrulle
(a) 3.0 (b) 1.5
¯(ln(d˜¶/¯d˜¶˘)˘, ß 2(ln(d˜¶ ))
2.5 1.0
2.0
0.5
C(h)
1.5
0
1.0
-0.5
0.5
0 -1.0
-0.2 0 0.2 0.4 0.6 0.8 -7 -6 -5 -4 -3 -2 -1
h ln(¶/L)
F IGURE 27. (a) Multi-fractal spectrum C(h) for the experiments ANTIG and ANTIC-1 to
ANTIC-4. The spectrum has been obtained by taking the inverse Legendre transform of
the scaling exponents ζ (p) shown in figure 13(a). The dotted line is a parabolic fit C(h) =
(h − a)2 /2b with a = 0.35 and b = 0.045. (b) Further test of the WKHM–multi-fractal
refined similarity RSH2, showing that the mean and the variance of ln(Dν` /hDν i) obey
the same scalings than the mean and variance of ln(` ). The dotted and dashed lines are
the same fits than in figure 16(b).
of ` (failure of K62), and looks closer to that of DI` . However, if we consider now
τrsh = ζ (p) − ζ (3)p/3, we observe a very good coincidence: this means that a correct
refined similarity hypotheses is
h|δW` |p i h|DI` |p/3 i
=
h|δW` |3 ip/3 h|DI` |ip/3
ξ I ( p/3)
`
= Cp (`) p/3
, (7.14)
L
where ξ I is given by (7.13), in the log-normal approximation for the MFR spectrum.
This is in agreement with the remark by Kraichnan (1975), since DI` is a quantity
that traces local energy transfers. The physical interpretation was given in § 7.3.1,
where we saw the connection between location of high energy transfers, and
dissipative solutions with h 6 1/3.
Does it mean that Kolmogorov’s intuition was wrong, and that the energy
dissipation is useless to characterize intermittency? In fact, no. If we observe
further figure 13(b), we see that the scaling exponents of ` and Dν` /hDν i coincide.
In fact, as shown in figure 27(b), the scaling properties of the mean and the variance
of ln(Dν` /hDν i) also coincide with that of ln(` ). This means that there is a refined
similarity hypothesis involving ` that states
D p E D ` p
` ν
= . (7.15)
hDν` i
867 P1-50
Beyond K41
This refined similarity hypothesis imposes ξ ν (p) − pξ ν (1) = τ (p) = µp(1 − p), or,
using (7.13), 2b = µ. This relation is indeed well satisfied, since µ = 0.13/2 = 0.065.
So we can indeed deduce the intermittency corrections from the measurements
of energy dissipation over a ball `, provided we use the new refined similarity
hypothesis (7.15).
8. Prospects associated with the multi-fractal WKHM picture of turbulence

The combination of the weak formulation of Kármán–Howarth–Monin equation
and multi-fractal theory (WKHM–MFR picture) provides us with a better understanding
of the turbulent cascade properties, and of the building of local large events of
energy transfer and dissipation. This raises a number of new issues and prospects,
a few of which are described below.
8.1. Link with coherent structures

After reading a preliminary version of this paper, Paul Linden asked me whether the
structures depicted in figure 22, had anything to do with the ‘coherent structures’
discussed by Jimenez (2018) (or – this is my personal addition – by Farge &
Schneider (2001)). It is not easy to answer this question because, as noted by
Jimenez (2018), there is no clear mathematical definition of ‘coherent structures’.
Typically, they appear as soon as one applies a threshold to a suitable quantity
(empirically, at least quadratic in the velocity, or involving spatial derivatives).
Farge & Schneider (2001) go a little bit further by suggesting to define them
using a threshold upon the orthogonal wavelet coefficients of the vorticity field.
With this definition, similar to a denoising procedure, the leftover appears as
an ‘incoherent noise’, with Gaussian statistics and a sheet-like topology, while the
coherent structures are being described by typically less than 5 % of the coefficients,
while keeping 90 % of the energy or enstrophy, and filament-like topology (Farge
& Schneider 2001). As pointed out by Farge & Schneider (2001), such a definition
enables the introduction of an interesting alternative to large-eddy (LES) or Reynolds
averaged (RANS) velocity field decomposition, where now the velocity field is split
into
u = uc + ui , (8.1)
where uc is the velocity field of the coherent structure, reconstructed by Bio–Savart
law applied to the denoised vorticity field, and ui = u − uc is the leftover.
In the case of figure 22, the structures appear indeed clearly after application of
a thresholding over the wavelet transform of a quantity that is cubic or quadratic in
velocity, and contains at least a space derivative. In that respect, it can be regarded
as a true ‘coherent structure’. However, it is not exactly equivalent to a ‘coherent
structure’ defined using the denoising procedure applied on the vorticity field.
867 P1-51
B. Dubrulle
(a) |øc| (b) |øc|

70
40
0.05 35 0.05 60
30 50
25 40
y 0 0
20
30
15
20
-0.05 10 -0.05
10
5
-0.05 0 0.05 -0.05 0 0.05

x x
F IGURE 28. Examples of coherent structure defined using the denoising procedure
of Farge & Schneider (2001) applied to the vorticity field in experiment CONC-3D.
Around the event, coded in colour, the velocity field, shown by white arrows display a
‘shock-like’ (a) or ‘spiral-like’ (b) coherent structure, to be compared with the structures
of figure 22. The figures are drawn using unpublished data collected by the EXPLOIT
collaboration (F. Daviaud, B. Dubrulle, P. Debue, V. Valori, J.-P. Laval, J.-M. Foucaut,
Ch. Cuvier, Y. Ostovan), with permission. Adapted from Debue (2019).
The corresponding coherent structure is provided in figure 28 for the same fields
as in figure 22. We see than the structure defined using vorticity and the structures
defined using DI` or Dν` do not exactly coincide, illustrating the arbitrariness of the
definition. On the other hand, they share many topological aspects. If I were asked
how to define a coherent structure function, I would personally advise the use of
the coherent structures computed using a threshold on DI` because they are clearly
associated with large energy transfer and potential quasi-singularities. Further, they
provide a clearer link to irreversibility and out of equilibrium in terms of Farge’s
decomposition (8.1), as I discuss now.
8.2. Back to irreversibility

As we have seen previously, the intermittency corrections are the hallmark of
quasi-singularities, which also provide a natural mechanism for the anomalous
dissipation. Since the latter is formally associated with a spontaneous breaking of
the time-reversal symmetry, it is interesting to look for possible closer connection
between the quasi-singularities and such symmetry breaking. A first natural step is
to decompose the velocity field into its odd and even components with respect to
time reversal, namely
u(x, t) = u+ + u− ,
u+ = 12 (u(x, t) + u(x, −t)) , (8.2)
u = (u(x, t) − u(x, −t)) .
− 1
2
(8.3)
867 P1-52
Beyond K41
From their symmetry through time reversal: u+ → u+ ; u− → −u− , it is clear
that u− represents the reversible part of the velocity, while u+ represents the
irreversible part of the velocity. Intuitively, we deduce that if there is any singularity
or quasi-singularity that contributes to the anomalous dissipation, it should be
contained in the irreversible component, u+ . On the other hand, since u− is
reversible, we expect it to represent some sort of equilibrium state. For example, if
we apply this decomposition to the finite viscosity solution of the Burgers equation
uν (x, t) = (x − L tanh(Lx/2νt))/t, we get
u+ = −L tanh(Lx/2νt)/t, (8.4)
x
u− = . (8.5)
t
Indeed, the singular behaviour as ν → 0 is contained in u+ , while u− is a large-scale
regular field for t > 0.
Now, it is interesting to rewrite the Navier–Stokes equations using the decomposition
(8.2) and (8.3). Keeping track of the symmetries under time reversal, we obtain a
set of two coupled equations
1 −
i = −uj ∂j ui −
Dt u+ ∂i p + ν∂j ∂j u−
i ,
+ −
(8.6)
ρ
1 +
i = −uj ∂j ui −
Dt u− ∂i p + ν∂j ∂j u+i ,
+ +
(8.7)
ρ
where Dt = ∂t + u− j ∂j represents the advection by the (regular) reversible component.
This set of equations formally resembles the set derived by Laval, Dubrulle &
Nazarenko (2001) using rapid distortion theory, with the role of the large-scale
flow being played by the reversible component, and that of the small scale being
played by the irreversible component. Laval et al. (2001) noticed that the small
scale is multiplicatively and additively forced by the large scale, and derived a
corresponding Fokker–Planck equation to derive the probability distribution for
the small-scale velocity increments. The result is a distribution with three salient
features: (i) algebraic tails at small scales, characteristic of a singular behaviour;
(ii) exponential cutoff at large scale; (iii) non-zero skewness induced by the coupling
between the multiplicative and additive noise.
These observations provide further support for the tempting idea to associate the
irreversible part of the process with small-scale quasi-singularities that are built by
stretching and advection of the regular, time reversible field, but additional work is
needed to strengthen this point.
8.3. Quasi-singularities of Navier–Stokes?

Local energy transfers have evidenced what could be the footprint of quasi-
singularities of the Euler equation (also corresponding to Onsager’s dissipative
867 P1-53
B. Dubrulle
solutions). Some work is needed to make the link between these observations,
and abstract analytical or numerical construction of such solutions using e.g. the
Mikado technique of Buckmaster et al. (2018). Other open question are: what is
the mechanism of formation of these large energy transfers? Are they linked to
self-similar blow-up solutions of the Euler or Navier–Stokes equations? This seems
to be true at least in shell model of turbulence, as proved by Mailybaev (2013).
Here, blow-up solutions only exist in the inviscid limit. In the finite viscosity case,
solutions representing regularized blow-up solutions generate coherent structures,
which travel through the inertial range in finite time and are described by universal
self-similar statistics. The intermittency can then be related analytically to the
process of instanton creation using a large deviation principle. Generalizing this
approach to turbulence is probably too hard at the present time, but interesting new
lines of research can be drawn using these ideas.
8.4. How deep do we have to dive into the dissipative range?

The extrapolation of measurements of the multi-fractal spectrum provided hmin ≈
−0.2. Whether this value is realistic and representative of the actual intermittency
can be a subject of debate, and is left to further research. What cannot be debated
is that hmin differs from 1/3 by more than 50 %. For sake of illustration, let us use
the value of hmin = −1/5 for discussion. One thing any student of turbulence learns
is that, to be resolved, a turbulence simulation must extend towards at least the
Kolmogorov scale. So, to save computational resources, it is tempting to choose a
resolution of 1x = η or kmax η = 1 for spectral codes. To capture the large events
of energy transfer and dissipation produced by quasi-singularities with Hölder
exponents around hmin , one must go at least to 1x = ηhmin ∼ Re5/4 ∼ ηRe−1/2 . For
a spectral DNS at Rλ = 36, this would mean it is necessary to go to kmax η = 6,
while it reaches kmax η = 12 for Rλ = 144. This is catastrophic for spectral numerical
simulation, since now the number of degrees of freedom to consider scales like
R15/4
λ instead of the ‘classical’ R9/4
λ law, based on the Kolmogorov scale. So instead
of using (128)3 modes for my Rλ = 36 simulation, I have to use (768)3 modes, and
my larger Reynolds simulation (512)3 requires now (6144)3 modes. Reaching good
statistics with these resolutions is beyond present computer capacity. Experiments
are still ahead for quite a good time in that respect.
Of course, a clever way out of this would be to increase the resolution only at the
points where h < 1/3. That way, we decrease very much the computational burden.
This is the idea behind adaptive-refined mesh programs like the open source Basilisk
code of Stéphane Popinet (to be found at http://basilisk.fr/). In my opinion, this
appears as the most promising road to get out of this numerical trap.
From the experimental point of view, decreasing the resolution in particle image
velocimetry is also problematic because of noise issues, and optics requirements.
With the best available cameras, and the best available data processing, it is
867 P1-54
Beyond K41
presently difficult to go much below 0.5 mm in resolution, for a fully turbulent
flow. For the von Kármán turbulent flow, the Kolmogorov scale varies like
η/R = 4 × 10−3 (Re/6000)−3/4 where R is the radius of the experiment, and the
flow is fully turbulent only above Re = 6000. For the 10 cm experiment in Saclay,
the maximal Kolmogorov scale we can reach in the turbulent regime is η = 0.4 mm
at Re = 6000 (Rλ = 214). So the only way to reach a fraction of η with present
optics is to increase the size of the experiment. For R = 1 m (already a huge
experiment!), we get η = 4 mm, so that we can hope to reach 1x = η/8, and
resolve events corresponding to h = −0.1. To reach h = hmin = −1/5, we would
need to reach 1x = η/14, only achieved in an experiment of size R = 7 m without
changing the optics. Clearly, some improvement in optics will be mandatory to
achieve our dream of resolving the highest dissipative events!
8.5. Why is turbulence log-normal? Or is it really?

During our journey to turbulence, we have met two occurrences of log-normality:
one in energy dissipation over a ball of size ` (within the framework of K62)
and one in the description of intermittency for velocity wavelet structure functions.
In § 5.3, we have argued that log-normality is a natural outcome of both scale
invariance and Kramer’s theorem, for fluctuations close to the mean. However,
in the WKHM–MFR picture, there is room for large fluctuations of local energy
transfer, and energy dissipation, so the ‘closeness to the mean’ hypothesis does
not hold, and any kind of probability distribution could a priori be expected. If
we only impose scale invariance as a constraint, we could for example expect
any log-infinitely divisible law (She & Waymire 1995), besides log-normal. So the
question of whether the latter is selected is a puzzle. There are some attempts to
connect this to the large-scale behaviour of the velocity increments by continuity
of the cascade (Castaing, Gagne & Hopfinger 1990; Boffetta et al. 2008). The
turbulence would then be log-normal because velocity increments are Gaussian at
large scale. But this means only postponing the problem, since it is not clear either
why turbulence is Gaussian at large scale.
One point of view could be that we do not see deviations from log-normality
because we did not go deep enough into the dissipative range: by staying close to
the Kolmogorov scale, we miss the large fluctuations arising from quasi-singularities
with h 1/3. So it would be interesting to increase the resolution of both the DNS
and the experimental measurements, but the price to pay is huge, see § 8.4.
If the turbulence is indeed log-normal, the understanding of intermittency can
be pinned to the computation of b from first principles or from the Navier–Stokes
equations. Field theory paves our way to a methodology that provides the solution
in Kraichnan’s model of turbulence (Falkovich et al. 2001). However, nobody
has succeeded to implement such methodology in INSE so far, and we lack
understanding of what fixes the value of b. It is well known that the value of
867 P1-55
B. Dubrulle
b is approximately two times smaller when one considers longitudinal velocity
increment (instead of the wavelet) based structure functions, which are closer to
transverse velocity increment based structure functions (Kestener & Arneodo 2004).
Can the value of b be fixed by simple geometrical arguments or by fixing the value
of C(h) for special values of h? For example, a loose interpretation of Cafarelli’s
theorem (Caffarelli et al. 1982) about the probability of singularity in Navier–Stokes
tells us that C(−1) > 4, which provides bounds on b. Maybe it would be possible
to derive a theorem regarding the probability of regular points in Navier–Stokes,
that would provide another bound on b via C(1) (see also § 8.9).
8.6. Subgrid modelling?

The K41 energy cascade picture is at the basis of a lot of subgrid modelling
of turbulence, to simulate flows at finite resolution 1x > η. For example, the
‘eddy-viscosity’ model relies on the idea that any energy transfer at a given scale `
will eventually results in an energy dissipation at the Kolmogorov scale. One thus
replaces the energy transfer by a viscous term, with a viscosity tuned to provide
an energy dissipation equal to the energy transfer. If 1x η, the velocity gradients
at this scale, of the order of 1u/1x are much lower that what they would be
at scale η. The turbulent viscosity is therefore increased by a huge factor (of the
order of (1x/η)2 ) with respect to the molecular viscosity. While this approach has
enabled a successful computation of large-scale flows in many fields of interest
(aeronautical, geophysics, astrophysics), it is not very appropriate to account for the
small-scale inhomogeneity and fluctuations that result from the true multi-fractal
local cascade. This is problematic in combustion or in problems involving phase
transition (like rain or cloud formation), since chemical reactions preferably occur
at locations of high energy dissipation, where due to energy release the temperature
locally increases. This cannot be captured by an eddy-viscosity model, that spreads
out homogeneously all dissipation over a cell of size 1x.
The issue here is to find a subgrid model that captures the localization and energy
dissipation focusing property evidenced by the WKHM–MFR approach. I know of
several approaches that appear promising in this context. The first one is to use
simple random multiplicative models, introduced originally by Mandelbrot (1972),
Meneveau & Sreenivasan (1991). Such models naturally localize and concentrate
energy dissipation. Then, by adjusting the probabilities, one can produce statistically
realistic energy dissipation events. A more realistic but more complex version of
this idea is to use multiplicative chaos (exponential of random matrices) to produce
synthetic small-scale velocity fields (Pereira, Garban & Chevillard 2016). A third
approach, developed by Nore et al. (2018) is to introduce an ‘entropy viscosity’,
that locally adds a diffusion depending on the unbalance in the energy equation at
scale 1x. This method is therefore ‘rooted’ in the WKHM picture, and produces a
viscosity that behaves differently depending on the local regularity of the flow: it is
867 P1-56
Beyond K41
smaller than the order of consistency of the method in regions where the solution
is smooth and well resolved and does not perturb the approximation. On the other
hand, in regions where the solution is not well approximated due to the presence of
large gradients that cannot be represented by the coarse mesh, the entropy viscosity
adds a diffusion proportional to the unbalance in the energy equation so the resulting
approximation dissipates energy.
8.7. Towards energetics of turbulence?

The investigation of local kinetic energy transfers has helped us to unveil
the properties of the energy cascade. It is not very difficult to generalize the
WKHM approach to other types of local energy transfer, such as thermal energy
transfers in convection (Faranda et al. 2018) or magnetic energy transfers in
magnetohydrodynamic turbulence or plasma physics (Galtier 2018). In these two
examples, there is actually an interplay between the different types of energy: in
Rayleigh–Bénard convection, for example, the energy is fed into the system as
thermal energy, and converted into kinetic energy under the action of buoyancy, to
be finally converted again into thermal energy by viscous dissipation. Such a global
cycle may have a very interesting local counter-part, that may be the main engine
governing the formation of coherent structures like plumes, or tornadoes. Clearly,
the extension of WKHM to this fields of geophysical or astrophysical interest will
provide much interesting information regarding the energetics and the dynamics of
such flows.
8.8. What is the origin of anomalous dissipation?

One motivation of the theory of K41 was to provide a self-consistent picture
of turbulence, including anomalous dissipation as a building block. Onsager’s
conjecture went one step further by suggesting that the anomalous dissipation
was connected to dissipative solutions of the Euler equations. The WKHM–MFR
formalism provides experimental support for this idea and shows that the anomalous
dissipation is a direct consequence of the 4/3th law of turbulence, via the constraints
it imposes on the multi-fractal spectrum C(h).
This view may not be fully satisfying for my colleagues of statistical physics, who
would like to link the anomalous dissipation with a spontaneous breaking of the
time-reversal symmetry of Navier–Stokes solutions with much more rigour than the
basic discussion provided in § 8.2. A way to content them would be to prove that
either the dissipative solutions originate from this time reversal symmetry breaking,
or that dissipative solutions induce time reversal symmetry breaking, by generating
for example multiple solutions. Results along these lines have already been obtained
by Eyink & Drivas (2015) in the Burgers case.
867 P1-57
B. Dubrulle
8.9. Conjecture: symmetry of the multi-fractal spectrum
Because of K41 theory, we have ζ I (1) = 0. By definition, we also have ζ I (0) = 0, so
that ζ I (1) = ζ I (0). A bold extrapolation leads us to conjecture that ζ I (1 − p) = ζ I (p)
for any p. Testing such a conjecture is not so easy: it requires good statistics and
precision both for large values of DI` (to get values of ζ I (p) for large p > 0) and for
small values of DI` (to get values of ζ I (p) for large p < 0). Experiments perform well
for the first requirement, but not for the second requirement (because of noise issues
that limit the precision), while numerical simulations perform well for the second
requirement (especially for spectral methods), but not for the first, because of limited
spatial resolution and statistics. This conjecture does not come out of the blue: it is
observed for example in field theories of Anderson localization or in the random
Potts model in two dimensions (Monthus, Berche & Chatelain 2009). It is in fact
closely connected to a possible symmetry of the multi-fractal spectrum, which by
Legendre transform can be readily seen to obey
C(h) − C(−h) = −3h, (8.8)
if the symmetry of ζ I holds. Remember that C is a large deviation function. This
kind of symmetry is quite classical in out-of-equilibrium large deviation theories
(Touchette 2009). As discussed by Monthus et al. (2009), the symmetry can be used
to discriminate between intermittency models. For example, it is straightforward to
show that the log-normal model (7.12) satisfies the symmetry (8.8), but not the
She–Lévêque model, given by (Boffetta et al. 2008)

2β − 3h − 1 2β − 3h − 1
C(h) = − 1 − ln , (8.9)
ln(β) 2 ln β
where β = 2/3. If the conjecture is correct, this means that the She–Lévêque model
cannot be valid in turbulence, a conclusion also reached by Granero-Belinchón, Roux
& Garnier (2018) using arguments based on information theory.
The conjecture, if valid, could also help finding proof of existence of singularities
in Navier–Stokes, since for h = 1, it means: C(−1) = 3 + C(1). So, we can get
information about the probability of observing the Navier–Stokes singularity by
looking at regular points and their probability! This is probably mathematically
easier to study.
8.10. What is the Lagrangian equivalent of WKHM?

The present more accurate measurements of velocity fields using imaging are based
on particle tracking, and provide information about particle trajectories, and the
velocity and acceleration along these trajectories. The study of their properties form
the so-called Lagrangian turbulence (Mordant et al. 2001; Biferale et al. 2004; Xu
et al. 2006). The multi-fractal theory of turbulence has already been generalized
867 P1-58
Beyond K41
to Lagrangian turbulence (Arneodo et al. 2008) and the Lagrangian and Eulerian
multi-fractal spectra have been unified in Chevillard et al. (2012). A natural question
is then: can we go further and unify also the WKHM description? This requires
the definition of Lagrangian local energy transfer, and the building of a Lagrangian
equivalent to the quantity DI` and Dν` . While I was finalizing the present essay, I
noticed that the problem has been recently solved by Drivas (2018), who establishes
Lagrangian formulae for energy conservation anomalies involving the discrepancy
between short-time two-particle dispersion forward and backward in time. These
results provide support to an initial work by Jucha et al. (2014) and rely on a
rigorous version of the Ott–Mann–Gawedzki relation (Ott & Mann 2000; Falkovich
et al. 2001), sometimes described as a ‘Lagrangian analogue of the 4/3-law’. It will
be interesting to dig further into these results, both theoretically and experimentally
or numerically, to understand all their implications.
Let me conclude by a last citation from Batterman (2011), that expresses perfectly
my opinion about the role of singularities in turbulence: many physicists and
philosophers apparently believe that singularities appearing in our theories are
indications of modelling failures [. . .]. Singularities are, in this view, information
sinks [. . .] On the contrary, I am suggesting that an important lesson from the
renormalization group successes is that we rethink the use of models in physics. If
we include mathematical features as essential parts of physical modelling then we
will see that blow ups or singularities are often sources of information.
Acknowledgements
This essay would not have been possible without the help of all my collaborators
and students. My special thanks go to J.-P. Laval, V. Shukla, F. Nguyen, H. Falller
and D. Geneste for providing the numerical data and analysis and to F. Daviaud,
A. Chiffaudel, J.-M. Foucaut, C. Cuvier, Y. Ostovan, V. Padilla, C. Wiertel,
P. Diribarne, P. Cortet, E. Herbert, D. Faranda, E.-W. Saw, V. Valori, R. Monchaux,
B. Saint-Michel, S. Thalabard, D. Kuzzay, P. Debue for the particle velocimetry
measurements and analysis, and allowing me to use unpublished data in figures 9,
22, 25 and 28. The torque measurements in water and glycerol owe much to the
work of F. Ravelet and L. Marié and were extended to a wide range of Reynolds
number thanks to a collaboration between the Saclay experimental team, the VKS
collaboration and the SHREK collaboration. I thank them all. I also extend a
special thanks to my colleagues of the VKS collaboration for allowing me to use
unpublished data in figure 14 and to Bernard Rousset for sharing with me his torque
analysis. H. Faller, P. Debue, C. Doering and P. Linden provided a critical reading
of this manuscript that helped me clarifying some crucial points. This work has been
supported by the ANR EXPLOIT, grant agreement no. ANR-16-CE06-0006-01.
867 P1-59
B. Dubrulle
References
AGAFONTSEV, D. S., K UZNETSOV, E. A. & M AILYBAEV, A. A. 2016 Development of high vorticity
in incompressible 3d Euler equations: influence of initial conditions. J. Expl Theor. Phys. Lett.
104 (10), 685–689.
AGAFONTSEV, D. S., K UZNETSOV, E. A. & M AILYBAEV, A. A. 2017 Asymptotic solution for
high-vorticity regions in incompressible three-dimensional Euler equations. J. Fluid Mech. 813,
R1.
A RNEODO , A., B AUDET, C., B ELIN , F., B ENZI , R., C ASTAING , B., C HABAUD , B., C HAVARRIA ,
R., C ILIBERTO , S., C AMUSSI , R., C HILL , F. et al. 1996 Structure functions in turbulence,
in various flow configurations, at Reynolds number between 30 and 5000, using extended
self-similarity. Europhys. Lett. 34 (6), 411–416.
A RNEODO , A., B ENZI , R., B ERG , J., B IFERALE , L., B ODENSCHATZ , E., B USSE , A., C ALZAVARINI ,
E., C ASTAING , B., C ENCINI , M., C HEVILLARD , L. et al. 2008 Universal intermittent properties
of particle trajectories in highly turbulent flows. Phys. Rev. Lett. 100, 254504.
BATTERMAN , R. W. 2011 Emergence, singularities, and symmetry breaking. Foundations Phys. 41
(6), 1031–1050.
B EC , J. & K HANIN , K. 2007 Burgers turbulence. Phys. Rep. 447 (1), 1–66.
B ENZI , R. & B IFERALE , L. 2009 Fully developed turbulence and the multifractal conjecture. J. Stat.
Phys. 135 (5), 977–990.
B ENZI , R., C ILIBERTO , S., T RIPICCIONE , R., B AUDET, C., M ASSAIOLI , F. & S UCCI , S. 1993
Extended self-similarity in turbulent flows. Phys. Rev. E 48, R29–R32.
B IFERALE , L., B OFFETTA , G., C ELANI , A., D EVENISH , B. J., L ANOTTE , A. & T OSCHI , F. 2004
Multifractal statistics of Lagrangian velocity and acceleration in turbulence. Phys. Rev. Lett.
93, 064502.
B OFFETTA , G., M AZZINO , A. & V ULPIANI , A. 2008 Twenty-five years of multifractals in fully
developed turbulence: a tribute to Giovanni Paladin. J. Phys. A: Math. Theoret. 41 (36),
363001.
B RADSHAW, Z. & T SAI , T.-P. 2018 Self-similar solutions to the Navier–Stokes equations: a survey of
recent results. In Nonlinear Analysis in Geometry and Applied Mathematics, Part 2, Harvard
CMSA Series in Mathematics, vol. 2, p. 159. International Press.
B UCKMASTER , T., D E L ELLIS , C., S ZÉKELYHIDI , L. & V ICOL , V. 2019 Onsager’s conjecture for
admissible weak solutions. Commun. Pure Appl. Maths 72 (2), 229–274.
C AFFARELLI , L., K OHN , R. & N IRENBERG , L. 1982 Partial regularity of suitable weak solutions of
the Navier–Stokes equations. Commun. Pure Appl. Maths 35 (6), 771–831.
C AMPOLINA , C. S. & M AILYBAEV, A. A. 2018 Chaotic blowup in the 3d incompressible Euler
equations on a logarithmic lattice. Phys. Rev. Lett. 121, 064501.
C ASTAING , B., G AGNE , Y. & H OPFINGER , E. J. 1990 Velocity probability density functions of high
Reynolds number turbulence. Physica D 46 (2), 177–200.
C HAE , D. 2007 Nonexistence of self-similar singularities for the 3D incompressible euler equations.
Commun. Math. Phys. 1 (273), 203–215.
C HEVILLARD , L. 2004 Description multifractale unifiée du phénomène d intermittence en turbulence
Eulérienne et Lagrangienne. PhD thesis, Université Sciences et Technologies - Bordeaux I.
C HEVILLARD , L., C ASTAING , B., A RNEODO , A., L ÉVÊQUE , E., P INTON , J.-F. & ROUX , S. G.
2012 A phenomenological theory of Eulerian and Lagrangian velocity fluctuations in turbulent
flows. C. Rend. Phys. 13 (9), 899–928; structures and statistics of fluid turbulence/Structures
et statistiques de la turbulence des fluides.
C LUSEL , M. & B ERTIN , E. 2008 Global fluctuations in physical systems: a subtle interplay between
sum and extreme value statistics. Intl J. Mod. Phys. B 22 (20), 3311–3368.
D EBUE , P. 2019 Experimental approach of the Euler and Navier–Stokes singularities problem. PhD
thesis, Université Paris-Saclay, Paris.
867 P1-60
Beyond K41
D EBUE , P., S HUKLA , V., K UZZAY, D., FARANDA , D., S AW, E.-W., DAVIAUD , F. & D UBRULLE ,
B. 2018 Dissipation, intermittency, and singularities in incompressible turbulent flows. Phys.
Rev. E 97, 053101.
D OMBRE , T. & G ILSON , J.-L. 1998 Intermittency, chaos and singular fluctuations in the mixed
Obukhov-Novikov shell model of turbulence.. Physica D 111 (1), 265–287.
D RIVAS , T. D. 2019 Turbulent cascade direction and Lagrangian time-asymmetry. J. Nonlinear Sci.
29, 65.
D RIVAS , T. D. & E YINK , G. L. 2017 An onsager singularity theorem for leray solutions of
incompressible Navier–Stokes. arXiv:1710.05205.
D UCHON , J. & ROBERT, R. 2000 Inertial energy dissipation for weak solutions of incompressible
Euler and Navier–Stokes equations. Nonlinearity 13 (1), 249–255.
E GGERS , J. & F ONTELOS , M. A. 2009 The role of self-similarity in singularities of partial differential
equations. Nonlinearity 22, R1–R44.
E YINK , G. L. 2007–2008 Turbulence Theory. Course notes, The Johns Hopkins University. Available
at: http://www.ams.jhu.edu/eyink/Turbulence/notes/.
E YINK , G. L. & D RIVAS , T. D. 2015 Spontaneous stochasticity and anomalous dissipation for Burgers
equation. J. Stat. Phys. 158 (2), 386–432.
E YINK , G. L. & S REENIVASAN , K. R. 2006 Onsager and the theory of hydrodynamic turbulence.
Rev. Mod. Phys. 78, 87–135.
FALKOVICH , G., G AWȨDZKI , K. & V ERGASSOLA , M. 2001 Particles and fields in fluid turbulence.
Rev. Mod. Phys. 73, 913–975.
FARANDA , D., L EMBO , V., I YER , M., K UZZAY, D., C HIBBARO , S., DAVIAUD , F. & D UBRULLE , B.
2018 Computation and characterization of local subfilter-scale energy transfers in atmospheric
flows. J. Atmos. Sci. 75 (7), 2175–2186.
FARGE , M. & S CHNEIDER , K. 2001 Coherent vortex simulation (cvs), a semi- deterministic turbulence
model using wavelets. Flow Turbul. Combust. 66, 393–426.
F RISCH , U. 1996 Turbulence, the legacy of A. N. Kolmogorov. Cambridge University Press.
F RISCH , U. 2016 The collective birth of multifractals. J. Phys. A: Math. Theoret. 49 (45), 451002.
F RISCH , U. & PARISI , G. 1985 On the singularity structure of fully developed turbulence. In
Turbulence and Predictability in Geophysical Fluid Dynamics and Climate Dynamics (ed.
M. Gil, R. Benzi & G. Parisi), pp. 84–88. Elsevier.
G ALTIER , S. 2018 On the origin of the energy dissipation anomaly in (Hall) magnetohydrodynamics.
J. Phys. A: Math. Theoret. 51 (20), 205501.
G RANERO -B ELINCHÓN , C., ROUX , S. G. & G ARNIER , N. B. 2018 Kullback-Leibler divergence
measure of intermittency: application to turbulence. Phys. Rev. E 97, 013107.
J IMENEZ , J. 2018 Coherent structures in wall-bounded turbulence. J. Fluid Mech. 842, P1.
J UCHA , J., X U , H., P UMIR , A. & B ODENSCHATZ , E. 2014 Time-reversal-symmetry breaking in
turbulence. Phys. Rev. Lett. 113, 054501.
K ESTENER , P. & A RNEODO , A. 2004 Generalizing the wavelet-based multifractal formalism to random
vector fields: application to three-dimensional turbulence velocity and vorticity data. Phys. Rev.
Lett. 93 (4), 044501.
K IMURA , Y. & M OFFATT, H. K. 2018 A tent model of vortex reconnection under Biot–Savart
evolution. J. Fluid Mech. 834, R1.
K OLMOGOROV, A. N. 1941 The local structure of turbulence in incompressible viscous fluids for
very large Reynolds number. Dokl. Akad. Nauk SSSR 30, 301–305.
K OLMOGOROV, A. N. 1962 A refinement of previous hypotheses concerning the local structure of
turbulence in a viscous incompressible fluid at high Reynolds number. J. Fluid Mech. 13,
82–85.
K RAICHNAN , R. H. 1974 On Kolmogorov’s inertial-range theories. J. Fluid Mech. 62 (2), 305–330.
K RAICHNAN , R. H. 1975 Remarks on turbulence theory. Adv. Math. 16, 305.
867 P1-61
B. Dubrulle
K UZZAY, D., S AW, E.-W., M ARTINS , F. J. W. A., FARANDA , D., F OUCAUT, J.-M., DAVIAUD , F. &
D UBRULLE , B. 2017 New method for detecting singularities in experimental incompressible
flows. Nonlinearity 30 (6), 2381–2402.
L AVAL , J.-P., D UBRULLE , B. & N AZARENKO , S. 2001 Nonlocality and intermittency in
three-dimensional turbulence. Phys. Fluids 13 (7), 1995–2012.
L EBERRE , M. & P OMEAU , Y. 2018 Recording of Leray-type singular events in a high speed wind
tunnel. arXiv:1801.01762.
L ERAY, J. 1934 Sur le mouvement d’un liquide visqueux emplissant l’espace. Acta Math. 63, 193248.
L I , Y., P ERLMAN , E., WAN , M., YANG , Y., B URNS , R., M ENEVEAU , C., B URNS , R., C HEN , S.,
S ZALAY, A. & E YINK , G. 2008 A public turbulence database cluster and applications to study
Lagrangian evolution of velocity increments in turbulence. J. Turbul. 9, N31.
M AILYBAEV, A. A. 2012 Renormalization and universality of blowup in hydrodynamic flows. Phys.
Rev. E 85, 066317.
M AILYBAEV, A. A. 2013 Blowup as a driving mechanism of turbulence in shell models. Phys.
Rev. E 87, 053011.
M ANDELBROT, B. B. 1972 Possible refinement of the lognormal hypothesis concerning the
distribution of energy dissipation in intermittent turbulence. In Statistical Models and Turbulence
(ed. M. Rosenblatt & C. Van Atta), pp. 333–351. Springer.
M ENEVEAU , C. & S REENIVASAN , K. R. 1991 The multifractal nature of turbulent energy dissipation.
J. Fluid Mech. 224, 429–484.
M ONTHUS , C., B ERCHE , B. & C HATELAIN , C. 2009 Symmetry relations for multifractal spectra at
random critical points. J. Stat. Mech. 2009 (12), P12002.
M ORDANT, N., M ETZ , P., M ICHEL , O. & P INTON , J.-F. 2001 Measurement of Lagrangian velocity
in fully developed turbulence. Phys. Rev. Lett. 87, 214501.
M UZY, J. F., B ACRY, E. & A RNEODO , A. 1991 Wavelets and multifractal formalism for singular
signals: application to turbulence data. Phys. Rev. Lett. 67 (25), 3515–3518.
NAZARENKO , S. V. & G REBENEV, V. N. 2017 Self-similar formation of the Kolmogorov spectrum
in the Leith model of turbulence. J. Phys. A: Math. Theoret. 50 (3), 035501.
N ECAS , J., RUZIICKA , M. & S VERAK , V. 1996 On Leray’s self-similar solutions of the Navier–Stokes
equations. Acta Math. 176 (2), 283–294.
N ORE , C., C ASTANON , Q. D., C APPANERA , L. & G UERMOND , J.-L. 2018 Numerical simulation of
the von Karman sodium dynamo experiment. J. Fluid Mech. 854, 164–195.
O NSAGER , L. 1949 Statistical hydrodynamics. Il Nuovo Cimento 6 (2), 279–287.
OTT, S. & M ANN , J. 2000 An experimental investigation of the relative diffusion of particle pairs in
three-dimensional turbulent flow. J. Fluid Mech. 422, 207–223.
PALADIN , G. & V ULPIANI , A. 1987 Anomalous scaling laws in multifractal objects. Phys. Rep. 156
(4), 147–225.
P EREIRA , R. M., G ARBAN , C. & C HEVILLARD , L. 2016 A dissipative random velocity field for
fully developed fluid turbulence. J. Fluid Mech. 794, 369–408.
P UMIR , A. & S IGGIA , E. D. 1992 Finite-time singularities in the axisymmetric three-dimension Euler
equations. Phys. Rev. Lett. 68, 1511–1514.
R AVELET, F., C HIFFAUDEL , A. & DAVIAUD , F. 2008 Supercritical transition to turbulence in an
inertially driven von Kármán closed flow. J. Fluid Mech. 601, 339–364.
S AINT-M ICHEL , B. 2013 Von Karman flow as a paradigm for non-equilibrium statistical physics,
Université Pierre et Marie Curie.
S AINT-M ICHEL , B., H ERBERT, E., S ALORT, J., B AUDET, C., B ON M ARDION , M., B ONNAY, P.,
B OURGOIN , M., C ASTAING , B., C HEVILLARD , L., DAVIAUD , F. et al. 2014 Probing quantum
and classical turbulence analogy in von Karman liquid helium, nitrogen, and water experiments.
Phys. Fluids 26 (12), 125109.
S AW, E. W., K UZZAY, D., FARANDA , D., G UITTONEAU , A., DAVIAUD , F., W IERTEL -G ASQUET, C.,
PADILLA , V. & D UBRULLE , B. 2016 Experimental characterization of extreme events of inertial
dissipation in a turbulent swirling flow. Nature Comm. 7, 12466.
867 P1-62
Beyond K41
S HE , Z.-S. & WAYMIRE , E. C. 1995 Quantized energy cascade and log-poisson statistics in fully
developed turbulence. Phys. Rev. Lett. 74, 262–265.
DA S ILVA , C. B. & P EREIRA , J. C. F. 2008 Invariants of the velocity-gradient, rate-of-strain, and
rate-of-rotation tensors across the turbulent/nonturbulent interface in jets. Phys. Fluids 20 (5),
055101.
TAYLOR , G. I. 1938 Production and dissipation of vorticity in a turbulent fluid. Proc. R. Soc. 164
(916), 15–23.
T OUCHETTE , H. 2009 The large deviation approach to statistical mechanics. Phys. Rep. 478 (1),
1–69.
X U , H., B OURGOIN , M., O UELLETTE , N. T. & B ODENSCHATZ , E. 2006 High order Lagrangian
velocity statistics in turbulence. Phys. Rev. Lett. 96, 024503.
Y EUNG , P. K., Z HAI , X. M. & S REENIVASAN , K. R. 2015 Extreme events in computational
turbulence. Proc. Natl Acad. Sci. USA 112 (41), 12633–12638.
867 P1-63

Beyond Kolmogorov Cascades

Uploaded by

Copyright:

Available Formats

You might also like

Beyond Kolmogorov Cascades

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Beyond Kolmogorov Cascades

Uploaded by

Copyright:

Available Formats

Downloaded from https://www.cambridge.org/core.

J. Fluid Mech. (2019), vol. 867, P1, doi:10.1017/jfm.2019.98

Beyond Kolmogorov cascades

The large-scale structure of many turbulent flows encountered in practical situations

‘I soon understood that there was little hope of developing a pure,

Case F Glycerol Re Rλ  η 1x Symbol

(a) (b) f2, C2

dE/dt, Pinj, Pdiss, (Pinj - Pdiss)

2.2. Velocity correlations and energy spectra

(a) 105 (b)

2.3. Universality of the spectrum

3. The K41 theory and its consequences

3.2. K41 as a phenomenology of cascade and energy balance in scale space

3.3. Universality of the energy budget in scale

3.4. Statistical properties of the energy cascade in scale space

4. K41 theory and symmetries: successes and limitations

(a) log (¶/˙) (b) 0 (b1)

100 101 102 103 100 102

(a) (b) 0.6

100 102 100 102

4.2. Rescaling symmetry breaking and generalized skewness

F IGURE 9. Non-dimensional vorticity magnitude (a) and PDF of non-dimensional vorticity

4.3. Energy cascade and inhomogeneity

4.4. Vortex stretching and 1/3 rescaling

4.5. Global-scale symmetry breaking and intermittency

We then compute the wavelet velocity increments as

(a) (a1) log10(¶/˙) (b) (b1)

∂W(u)(x, ¶)* ∂W(u)(x, ¶)C*

100 102 100 102

100 101 102 103 100 101 102 103

(a) 3.0 (b)

≈˜( p/3), ≈ I( p/3)

4.6. Time-reversal breaking and enstrophy blow up

102 104 106 108

F IGURE 14. Non-dimensional energy injection rate /2L2 Ω 3 as a function of Reynolds

the Re = ∞ limit as proved by recent experiments using super-fluid helium 4 at

5. The Kolmogorov–Obukhov refined similarity hypothesis

5.2. Test of refined similarity hypothesis

ν (x, t) = ν (∂u)2 (x, t). (5.8)

(a) ln(¶/L) (b) 1

F IGURE 15. Probability distribution functions of the pseudo-energy dissipation averaged

(a) 1.5 (b) 2

5.3. Problems and issues set by K62

(a) 5 (b) 1.0

F IGURE 17. Test of refined similarity hypothesis. (a) Structure functions of ` as a

6. Weak formulation of Kármán–Howarth–Monin equation

6.1.2. How to implement a weak formulation?

The properties of φ guarantee that when ` → 0, u`i → ui . Moreover, the velocity

as well as any higher-order derivative of ui , as long as φ is sufficiently regular.

6.1.3. Advantages of the weak formulation

6.1.4. Summary of Duchon–Robert derivation

Developing terms and using properties of φ, it is simple to check that E ` can be

6.2. Properties of WKHM

6.2.3. Small-scale limit

At all points x where u is regular, we have δξ u = ξ · ∂u + O((ξ )2 ), so that for

where ν = ν(∂u)2 is the local viscous dissipation (potentially infinite if ∂u is

6.2.4. Inertial dissipation and Onsager’s conjecture

6.2.5. Illustration: inertial dissipation by a shock

Case F Glycerol Re Rλ η 1x Symbol

F IGURE 14. Non-dimensional energy injection rate /2L2 Ω 3 as a function of Reynolds

ν (x, t) = ν (∂u)2 (x, t). (5.8)

F IGURE 17. Test of refined similarity hypothesis. (a) Structure functions of ` as a

where ν = ν(∂u)2 is the local viscous dissipation (potentially infinite if ∂u is

= Dνηh = DIηh . (7.6)